EP3337135A1 - Acceleration management node, acceleration node, client, and method - Google Patents
Acceleration management node, acceleration node, client, and method Download PDFInfo
- Publication number
- EP3337135A1 EP3337135A1 EP16850315.9A EP16850315A EP3337135A1 EP 3337135 A1 EP3337135 A1 EP 3337135A1 EP 16850315 A EP16850315 A EP 16850315A EP 3337135 A1 EP3337135 A1 EP 3337135A1
- Authority
- EP
- European Patent Office
- Prior art keywords
- acceleration
- target
- acceleration device
- bandwidth
- node
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 230000001133 acceleration Effects 0.000 title claims abstract description 1221
- 238000000034 method Methods 0.000 title claims abstract description 65
- 238000007726 management method Methods 0.000 claims description 144
- 230000006870 function Effects 0.000 claims description 33
- 238000004364 calculation method Methods 0.000 claims description 14
- 230000002159 abnormal effect Effects 0.000 claims description 10
- 238000005516 engineering process Methods 0.000 description 8
- 238000010586 diagram Methods 0.000 description 7
- 238000004891 communication Methods 0.000 description 5
- 230000006835 compression Effects 0.000 description 2
- 238000007906 compression Methods 0.000 description 2
- 230000006837 decompression Effects 0.000 description 2
- 230000002093 peripheral effect Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 241000283973 Oryctolagus cuniculus Species 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000002035 prolonged effect Effects 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/08—Configuration management of networks or network elements
- H04L41/085—Retrieval of network configuration; Tracking network configuration history
- H04L41/0853—Retrieval of network configuration; Tracking network configuration history by actively collecting configuration information or by backing up configuration information
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/12—Protocols specially adapted for proprietary or special-purpose networking environments, e.g. medical networks, sensor networks, networks in vehicles or remote metering networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/44—Arrangements for executing specific programs
- G06F9/455—Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
- G06F9/45533—Hypervisors; Virtual machine monitors
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
- G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/08—Configuration management of networks or network elements
- H04L41/0896—Bandwidth or capacity management, i.e. automatically increasing or decreasing capacities
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/50—Network service management, e.g. ensuring proper service fulfilment according to agreements
- H04L41/5003—Managing SLA; Interaction between SLA and QoS
- H04L41/5019—Ensuring fulfilment of SLA
- H04L41/5025—Ensuring fulfilment of SLA by proactively reacting to service quality change, e.g. by reconfiguration after service quality degradation or upgrade
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/50—Network service management, e.g. ensuring proper service fulfilment according to agreements
- H04L41/5041—Network service management, e.g. ensuring proper service fulfilment according to agreements characterised by the time relationship between creation and deployment of a service
- H04L41/5051—Service on demand, e.g. definition and deployment of services in real time
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L65/00—Network arrangements, protocols or services for supporting real-time applications in data packet communication
- H04L65/40—Support for services or applications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/10—Protocols in which an application is distributed across nodes in the network
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/10—Protocols in which an application is distributed across nodes in the network
- H04L67/1001—Protocols in which an application is distributed across nodes in the network for accessing one among a plurality of replicated servers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2212/00—Indexing scheme relating to accessing, addressing or allocation within memory systems or architectures
- G06F2212/25—Using a specific main memory architecture
- G06F2212/254—Distributed memory
- G06F2212/2542—Non-uniform memory access [NUMA] architecture
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/50—Network services
- H04L67/60—Scheduling or organising the servicing of application requests, e.g. requests for application data transmissions using the analysis and optimisation of the required network resources
Definitions
- the present invention relates to the field of virtualization technologies, and in particular, to an acceleration management node, an acceleration node, a client, and a method.
- Hardware acceleration devices include a processor that provides particular instructions and other peripheral component interconnect (peripheral component interconnect, PCI) devices that can provide an acceleration function, such as a graphics processing unit (graphics processing unit, GPU) and a field programmable gate array (field programmable gate array, FPGA).
- PCI peripheral component interconnect
- Network Function Virtualization Network Function Virtualization
- NFV Network Function Virtualization
- a network device based on special-purpose hardware may be deployed on a general-purpose server by using a virtualization technology, to evolve from a conventional combination form of "embedded software + special-purpose hardware" to a combination form of "software + general-purpose hardware”.
- NF Network Function
- VNF Virtualization Network Function
- Embodiments of the present invention provide an acceleration management node, an acceleration node, a client, and a method, applied to a virtualization scenario, so that an acceleration device can be accurately invoked according to a service requirement of a client.
- an embodiment of the present invention provides an acceleration management node, where the acceleration management node includes: a receiving unit, configured to separately receive acceleration device information of all acceleration devices of each of at least one acceleration node reported by the acceleration node, where each acceleration node includes at least one acceleration device, and the acceleration device information includes an acceleration type and an algorithm type; an obtaining unit, configured to obtain an invocation request sent by a client, where the invocation request is used to invoke an acceleration device to accelerate a service of the client, and the invocation request includes a target acceleration type and a target algorithm type; an allocation unit, configured to query the acceleration device information to determine, from all the acceleration devices of the at least one acceleration node, a target acceleration device matching the invocation request; and an instruction unit, configured to instruct a target acceleration node on which the target acceleration device is located to respond to the invocation request.
- the acceleration management node invokes an acceleration device according to acceleration device information in each acceleration node and may allocate, according to a requirement of an application program of a client and an acceleration type and an algorithm type of each acceleration device, a corresponding acceleration device to the application program, so as to meet a service requirement.
- the allocation unit is configured to query the acceleration device information to determine, from all the acceleration devices of the at least one acceleration node, the target acceleration device whose acceleration type and algorithm type are respectively the same as the target acceleration type and the target algorithm type.
- the acceleration management node When invoking an acceleration device according to acceleration device information in each acceleration node, the acceleration management node performs invocation according to an acceleration type, an algorithm type, and acceleration bandwidth of each acceleration device, so as to ensure that bandwidth of the acceleration device can meet a service requirement, thereby implementing accurate invocation.
- the acceleration information further includes acceleration bandwidth, and the acceleration bandwidth includes total bandwidth and occupied bandwidth;
- the invocation request further includes target acceleration bandwidth;
- the allocation unit is specifically configured to query the acceleration device information to determine at least one candidate acceleration device whose acceleration type and algorithm type are respectively the same as the target acceleration type and the target algorithm type and whose remaining bandwidth is greater than or equal to the target acceleration bandwidth, and determine one of the at least one candidate acceleration device as the target acceleration device, where the remaining bandwidth is obtained by calculation according to the total bandwidth and the occupied bandwidth.
- the acceleration device information further includes non-uniform memory access architecture NUMA information;
- the invocation request further includes target NUMA information;
- the allocation unit is specifically configured to query the acceleration device information to determine at least one candidate acceleration device whose acceleration type and algorithm type are respectively the same as the target acceleration type and the target algorithm type, whose remaining bandwidth is greater than or equal to the target acceleration bandwidth, and whose NUMA information is consistent with the target NUMA information, and determine one of the at least one candidate acceleration device as the target acceleration device, where the remaining bandwidth is obtained by calculation according to the total bandwidth and the occupied bandwidth.
- the allocation unit is specifically configured to: when there is one candidate acceleration device, determine the candidate acceleration device as the target acceleration device.
- the allocation unit is specifically configured to: when there is a plurality of candidate acceleration devices, determine a first acceleration device having maximum remaining bandwidth from the plurality of candidate acceleration devices according to the acceleration bandwidth, and if there is one first acceleration device, determine the first acceleration device as the target acceleration device.
- the allocation unit is specifically configured to: when there is a plurality of first acceleration devices having the maximum remaining bandwidth, determine a second acceleration device having a maximum VF quantity from the plurality of first acceleration devices according to the VF quantity, and if there is one second acceleration device, use the second acceleration device as the target acceleration device.
- the allocation unit is specifically configured to: when there is a plurality of second acceleration devices having the maximum VF quantity, use a second acceleration device first found as the target acceleration device according to a time sequence of querying the acceleration device information.
- the instruction unit is specifically configured to send configuration instruction information to the target acceleration node, to instruct the target acceleration node to respond to the invocation request, where the configuration instruction message indicates an acceleration type and an algorithm type of the target acceleration device matching the invocation request, or the configuration instruction message indicates an acceleration type, an algorithm type, and acceleration bandwidth of the target acceleration device.
- the acceleration management node further includes a storage unit, configured to store the acceleration device information.
- the storage unit is further configured to: update previously stored acceleration device information corresponding to the target acceleration device according to the target acceleration bandwidth; and record an allocation result of the instruction unit.
- the acceleration management node further includes: a releasing unit, configured to obtain a release request sent by the client for releasing the target acceleration device, and invoke the target acceleration node to release the target acceleration device.
- the releasing unit is configured to: when detecting that the service of the client becomes abnormal, find the target acceleration device according to the allocation result recorded by the storage unit, and invoke the target acceleration node to release the target acceleration device.
- the storage unit is further configured to set the allocation result to invalid.
- an embodiment of the present invention provides an acceleration node, including: an agent unit, a driver, and at least one acceleration device, where the driver is configured to drive the at least one acceleration device, the at least one acceleration device is configured to provide a hardware acceleration function, and the agent unit is configured to: invoke the driver to separately query the at least one acceleration device to obtain acceleration device information of each acceleration device, where the acceleration device information includes an acceleration type and an algorithm type; and report the acceleration device information to an acceleration management node.
- the acceleration node reports acceleration device information of its acceleration devices to the acceleration management node, so that the acceleration management node can configure an appropriate acceleration device for a client according to the reported acceleration device information, thereby meeting a service requirement and implementing accurate invocation.
- the agent unit is further configured to: acceleration bandwidth.
- the agent unit is further configured to: receive a configuration instruction message sent by the acceleration management node, where the configuration instruction message indicates a target acceleration type and a target algorithm type of a target acceleration device matching an invocation request of a client, or the configuration instruction message indicates a target acceleration type, a target algorithm type, and target acceleration bandwidth of a target acceleration device matching an invocation request of a client; invoke, according to the configuration instruction message, the driver to detect whether the target acceleration device works normally; and when the target acceleration device works normally, configure a target interface of the target acceleration device for the client.
- the acceleration device information further includes non-uniform memory access architecture NUMA information.
- the configuration instruction message further indicates target NUMA information matching the invocation request of the client.
- the agent unit is further configured to configure the target acceleration type, the target algorithm type, and the target acceleration bandwidth as a hardware attribute of the target interface.
- the agent unit is further configured to respond to the acceleration management node and release the target acceleration device.
- the agent unit is further configured to set the hardware attribute of the target interface to null.
- an embodiment of the present invention provides a client, including: a requesting unit, configured to generate an invocation request according to an acceleration requirement of a service, where the invocation request includes a target acceleration type and a target algorithm type that are required for accelerating the service; and a sending unit, configured to send the invocation request to an acceleration management node to request to invoke a target acceleration device matching the invocation request to accelerate the service.
- An application program of the client may send, to the acceleration management node, a target acceleration type and a target algorithm type that correspond to an acceleration device that meets a requirement of accelerating a service of the application program, to apply to the acceleration management node for the acceleration device, so that the acceleration management node can more accurately invoke a target acceleration device required by the client.
- the invocation request further includes target acceleration bandwidth required for accelerating the service.
- the invocation request further includes target NUMA information required by the service.
- the requesting unit is further configured to: when the client needs to release the acceleration device, generate a release request for releasing the target acceleration device; and the sending unit is further configured to send the release request to the acceleration management node.
- an embodiment of the present invention provides an acceleration management method, including: separately receiving acceleration device information of all acceleration devices of each of at least one acceleration node reported by the acceleration node, where each acceleration node includes at least one acceleration device, and the acceleration device information includes an acceleration type and an algorithm type; obtaining an invocation request sent by a client, where the invocation request is used to invoke an acceleration device to accelerate a service of the client, and the invocation request includes a target acceleration type and a target algorithm type; querying the acceleration device information to determine, from all the acceleration devices of the at least one acceleration node, a target acceleration device matching the invocation request; and instructing a target acceleration node on which the target acceleration device is located to respond to the invocation request.
- an acceleration management node may allocate, according to a requirement of an application program of a client and an acceleration type and an algorithm type of each acceleration device, a corresponding acceleration device to the application program, thereby ensuring normal running of the accelerated service and implementing accurate invocation.
- the step of querying the acceleration device information to determine, from all the acceleration devices of the at least one acceleration node, a target acceleration device matching the invocation request includes: querying the acceleration device information to determine the target acceleration device whose acceleration type and algorithm type are respectively the same as the target acceleration type and the target algorithm type.
- the acceleration information further includes acceleration bandwidth, and the acceleration bandwidth includes total bandwidth and occupied bandwidth;
- the invocation request further includes target acceleration bandwidth;
- the step of querying the acceleration device information to determine, from all the acceleration devices of the at least one acceleration node, a target acceleration device matching the invocation request includes: querying the acceleration device information to determine at least one candidate acceleration device whose acceleration type and algorithm type are respectively the same as the target acceleration type and the target algorithm type and whose remaining bandwidth is greater than or equal to the target acceleration bandwidth, and determine one of the at least one candidate acceleration device as the target acceleration device, where the remaining bandwidth is obtained by calculation according to the total bandwidth and the occupied bandwidth.
- the acceleration device information further includes non-uniform memory access architecture NUMA information;
- the invocation request further includes target NUMA information;
- the step of querying the acceleration device information to determine, from all the acceleration devices of the at least one acceleration node, a target acceleration device matching the invocation request specifically includes: querying the acceleration device information to determine at least one candidate acceleration device whose acceleration type and algorithm type are respectively the same as the target acceleration type and the target algorithm type, whose remaining bandwidth is greater than or equal to the target acceleration bandwidth, and whose NUMA information is consistent with the target NUMA information, and determine one of the at least one candidate acceleration device as the target acceleration device, where the remaining bandwidth is obtained by calculation according to the total bandwidth and the occupied bandwidth.
- a fourth possible implementation manner when there is one candidate acceleration device, the candidate acceleration device is determined as the target acceleration device; or when there is a plurality of candidate acceleration devices, a first acceleration device having maximum remaining bandwidth is determined from the plurality of candidate acceleration devices according to the acceleration bandwidth, and if there is one first acceleration device, the first acceleration device is determined as the target acceleration device bandwidth.
- the acceleration device information further includes a virtual function VF quantity; and when there is a plurality of first acceleration devices having the maximum remaining bandwidth, a second acceleration device having a maximum VF quantity is determined from the plurality of first acceleration devices, and if there is one second acceleration device, the second acceleration device is used as the target acceleration device.
- a second acceleration device first found is used as the target acceleration device according to a time sequence of querying the acceleration device information.
- the method further includes: storing the acceleration device information.
- the method further includes: updating previously stored acceleration device information corresponding to the target acceleration device according to the target acceleration bandwidth; and recording an allocation result.
- the method further includes: obtaining a release request sent by the client for releasing the target acceleration device, and invoking the target acceleration node to release the target acceleration device.
- the method further includes: when detecting that the service of the client becomes abnormal, finding the target acceleration device according to the recorded allocation result, and invoking the target acceleration node to release the target acceleration device.
- the method further includes: setting the allocation result to invalid.
- an embodiment of the present invention provides an acceleration device configuration method, applied to an acceleration node, where the acceleration node includes a driver and at least one acceleration device; and the method includes: invoking the driver to separately query the at least one acceleration device to obtain acceleration device information of each acceleration device, where the acceleration device information includes an acceleration type and an algorithm type; and reporting the acceleration device information to an acceleration management node.
- the acceleration node reports acceleration device information of its acceleration devices to the acceleration management node, so that the acceleration management node can configure an appropriate acceleration device for a client according to the reported acceleration device information, thereby meeting a service requirement and implementing accurate invocation.
- the acceleration device information further includes acceleration bandwidth.
- the method further includes: receiving a configuration instruction message sent by the acceleration management node, where the configuration instruction message indicates a target acceleration type and a target algorithm type of a target acceleration device matching an invocation request of a client, or the configuration instruction message indicates a target acceleration type, a target algorithm type, and target acceleration bandwidth of a target acceleration device matching an invocation request of a client; invoking, according to the configuration instruction message, the driver to detect whether the target acceleration device works normally; and when the target acceleration device works normally, configuring a target interface of the target acceleration device for the client.
- the acceleration device information further includes non-uniform memory access architecture NUMA information.
- the configuration instruction message further indicates target NUMA information matching the invocation request of the client.
- the method further includes: configuring the target acceleration type, the target algorithm type, and the target acceleration bandwidth as a hardware attribute of the target interface.
- the method further includes: responding to the acceleration management node and releasing the target acceleration device.
- the method further includes: setting the hardware attribute of the target interface to null.
- an embodiment of the present invention provides a method of applying for an acceleration device, including: generating an invocation request according to an acceleration requirement of a service, where the invocation request includes a target acceleration type and a target algorithm type that are required for accelerating the service; and sending the invocation request to an acceleration management node to request to invoke a target acceleration device matching the invocation request to accelerate the service.
- An application program of a client may send, to the acceleration management node, a target acceleration type and a target algorithm type that correspond to an acceleration device that meets a requirement of accelerating a service of the application program, to apply to the acceleration management node for the acceleration device, so that the acceleration management node can more accurately invoke a target acceleration device required by the client, and ensure normal running of the service.
- the invocation request further includes target acceleration bandwidth required for accelerating the service.
- the invocation request further includes target NUMA information required by the service.
- the method further includes: when the acceleration of the service is completed, generating a release request for releasing the target acceleration device; and sending the release request to the acceleration management node.
- an embodiment of the present invention provides an acceleration management system, including: the acceleration management node according to any one of the first aspect or the first to thirteenth possible implementation manners of the first aspect and the acceleration node according to any one of the second aspect or the first to seventh possible implementation manners.
- the acceleration management system can accurately invoke, according to a service requirement of an application program of a client, an appropriate acceleration device to accelerate a service of the application program and ensure normal running of the service.
- Virtualization refers to virtualizing one computer into a plurality of logical computers by using a virtualization technology.
- a computation management module (not shown) in a client 300 (also referred to as a computer or a physical host) may create at least one virtual machine according to a user requirement.
- FIG. 1 shows three virtual machines in the client 300, namely, a virtual machine 300a, a virtual machine 300b, and a virtual machine 300c.
- the virtual machines may run on different operating systems.
- Each virtual machine may be considered as a logical computer.
- an acceleration node 200 may include general-purpose computer hardware, that is, an acceleration device such as a CPU or a GPU, and each virtual machine in the client 300 may invoke the acceleration device in the acceleration node 200 by using an acceleration management node 100.
- Network function virtualization (Network Function Virtualization, NFV) uses general-purpose hardware such as x86 and a virtualization technology to implement software processing of a lot of functions, so as to reduce network device costs. NFV may be considered as an actual application of the virtualization technology.
- the virtualization technology may also be applied in scenarios such as a public cloud, a private cloud, an enterprise cloud, and cloud acceleration. Therefore, the solutions of the embodiments of the present invention are not limited to an NFV scenario, and the protection scope of the present invention should not be limited thereto.
- FIG. 2 is a schematic structural diagram of an acceleration management node 100 according to an embodiment of the present invention.
- the acceleration management node 100 includes: a receiving unit 101, an obtaining unit 103, an allocation unit 104, and an instruction unit 105.
- the receiving unit 101 is configured to separately receive acceleration device information of all acceleration devices of each of at least one acceleration node reported by the acceleration node.
- Each acceleration node includes at least one acceleration device.
- the acceleration device information includes an acceleration type and an algorithm type.
- the acceleration type indicates a type of an acceleration function supported by each acceleration device.
- common acceleration functions may include encryption, decryption, compression, decompression, audio and video encoding and decoding, and the like.
- the algorithm type indicates an algorithm used by each acceleration device when the acceleration device implements an acceleration function supported by the acceleration device.
- FIG. 2 shows three acceleration nodes, namely, an acceleration node 200a, an acceleration node 200b, and an acceleration node 200c.
- An actual quantity of acceleration nodes may be set according to a network requirement. No limitation is imposed herein.
- Each acceleration node may include at least one hardware acceleration device such as a CPU, a GPU, or a PCI device.
- Each acceleration device has its own acceleration type, algorithm type, and the like. That is, each acceleration device corresponds to one piece of acceleration device information. Different acceleration devices may correspond to same or different acceleration device information.
- the acceleration node may separately report acceleration device information of the CPU and acceleration device information of the GPU to the acceleration management node 100 by using the receiving unit 101.
- the receiving unit 101 may be a reporting interface and may specifically include a software interface.
- the acceleration node 200a may invoke the reporting interface in a Remote Procedure Call Protocol (Remote Procedure Call Protocol, RPC) manner, to report the acceleration device information of all the acceleration devices in the acceleration node 200a to the acceleration management node 100.
- RPC Remote Procedure Call Protocol
- the other acceleration nodes 200b and 200c are similar to the acceleration node 200a, and details are not described herein again.
- the acceleration management node 100 in this embodiment of the present invention may be a management program running on a physical host.
- the physical host may usually include a processor, a memory, and an input/output (I/O) interface.
- the management program is stored in the memory.
- the processor can read and run the management program in the memory.
- the receiving unit 101 may be a software I/O interface, and the acceleration node may use various communications tools (for example, a communications tool Rabbit MQ) between software I/O interfaces to remotely invoke the software I/O interface for communication. It should be known by persons skilled in the art that communication may also be performed between software I/O interfaces by using various other message queues. No specific limitation is imposed herein.
- the obtaining unit 103 is configured to obtain an invocation request sent by a client 300.
- the invocation request is used to invoke an acceleration device to accelerate a service of the client 300, and the invocation request includes a target acceleration type and a target algorithm type.
- the client 300 may be specifically a physical host that runs an application program.
- the application program of the client 300 may notify, by using the obtaining unit 103, the acceleration management node 100 of information, such as a target acceleration type and a target algorithm type, of an acceleration device required for accelerating the service, so as to apply to the acceleration management node 100 for a hardware resource (that is, an acceleration device) for acceleration.
- the obtaining unit 103 may also be an application programming interface (application programming interface, API).
- the application program of the client 300 may communicate with the obtaining unit 103 by invoking an application programming interface. It should be known that each type of service of the application program requires a target acceleration type and a target algorithm type that comply with its service specification.
- the allocation unit 104 is configured to query the acceleration device information to determine, from all the acceleration devices of the at least one acceleration node, a target acceleration device matching the invocation request.
- the allocation unit 104 may search acceleration device information of all acceleration nodes that is stored in a storage unit 102, for a target acceleration device that meets the target acceleration type and the target algorithm type required by the invocation request.
- the instruction unit 105 is configured to instruct a target acceleration node on which the target acceleration device is located to respond to the invocation request and configure the target acceleration device for the client 300.
- the acceleration management node 100 further includes: the storage unit 102, configured to the acceleration device information.
- the storage unit 102 may store the obtained acceleration device information in a local memory of the acceleration management node 100 or in a network memory connected to the acceleration management node 100 through a network. No limitation is imposed herein.
- the storage unit 102 may store the acceleration device information in a list form, a database form, or another storage form well-known to persons skilled in the art.
- the acceleration management node 100 determines, by querying the acceleration device information obtained from the acceleration nodes 200a, 200b, and 200c, that a target acceleration device required for meeting the invocation request of the client 300 is located on the acceleration node 200c.
- the instruction unit 105 may invoke a configuration interface of the acceleration node 200c in an RPC manner to allocate the target acceleration type and the target algorithm type to the acceleration node 200c, so that the acceleration node 200c configures the corresponding target acceleration device for the client 300, thereby providing a hardware acceleration function for the service of the application program of the client 300.
- the acceleration management node invokes an acceleration device according to acceleration device information in each acceleration node and may allocate, according to a requirement of an application program of a client and an acceleration type and an algorithm type of each acceleration device, a corresponding acceleration device to the application program, so as to meet a service requirement.
- acceleration refers to allocating some services of an application program to a hardware acceleration device for operation. Because efficiency of logical operations of the hardware acceleration device is higher than that of a software algorithm, an operation time can be saved, thereby achieving acceleration.
- a specific acceleration type and algorithm type supported by the acceleration device are not considered.
- an encryption and decryption type such as IPSec
- the encryption and decryption type IPSec further includes three sub-types, namely, 3DES (Triple Data Encryption Algorithm, Triple Data Encryption Algorithm), DH (Diffie-Hellman Algorithm, D-H algorithm), and AES (Advanced Encryption Standard, Advanced Encryption Standard).
- 3DES Triple Data Encryption Algorithm
- Triple Data Encryption Algorithm Triple Data Encryption Algorithm
- DH Diffie-Hellman Algorithm, D-H algorithm
- AES Advanced Encryption Standard, Advanced Encryption Standard
- an acceleration device of the IPSec-DH type may be invoked for a service that requires an acceleration device of the IPSec-3DES type, resulting in that an invocation result cannot meet the requirement of the service.
- the technical solutions of the present invention can ensure accuracy of an invocation result, so that an attribute of an acceleration device invoked by the client 300 can meet a requirement of an application program.
- the allocation unit 104 may be specifically configured to query the acceleration device information to determine, from all the acceleration devices of the at least one acceleration node, the target acceleration device whose acceleration type and algorithm type are respectively the same as the target acceleration type and the target algorithm type.
- the acceleration information obtained by the receiving unit 101 may further include acceleration bandwidth.
- the acceleration bandwidth may include total bandwidth of an acceleration device and details of occupied bandwidth of the acceleration device at a current moment.
- the total bandwidth of the acceleration device refers to maximum acceleration bandwidth that can be provided by the acceleration device in a zero-load state.
- the invocation request obtained by the obtaining unit 103 may further include target acceleration bandwidth.
- the target acceleration bandwidth indicates bandwidth required by a service for which acceleration is requested by the client 300.
- the allocation unit 104 may first query the acceleration device information to obtain at least one candidate acceleration device whose acceleration type and algorithm type are respectively the same as the target acceleration type and the target algorithm type and whose remaining bandwidth is greater than or equal to the target acceleration bandwidth, that is, to obtain at least one candidate acceleration device matching the invocation request; and determine one of the at least one candidate acceleration device as the target acceleration device.
- the remaining bandwidth is obtained by calculation according to the total bandwidth and the occupied bandwidth.
- the acceleration management node 100 may allocate a corresponding acceleration device to the application program according to acceleration bandwidth of each acceleration device, so as to ensure that the allocated acceleration device can provide sufficient acceleration bandwidth for the application program, thereby implementing accurate invocation.
- the acceleration bandwidth of an acceleration device is not considered, and when remaining bandwidth of the acceleration device does not meet a service requirement, a prolonged acceleration time or an acceleration failure may also be caused, failing to achieve acceleration.
- the acceleration node 200 may provide a separate memory for each processor by using a non-uniform memory access architecture (Non Uniform Memory Access Architecture, NUMA), thereby avoiding performance loss caused by access of multiple processors to a same memory. Therefore, processors in the acceleration node 200 may be grouped according to NUMA information, and the processors in different groups have different NUMA information.
- the acceleration devices in the acceleration node 200 usually belong to different processors. Therefore, each acceleration device has same NUMA information as that of the processor to which the acceleration device belongs.
- the acceleration device information obtained by the receiving unit 101 may further include NUMA information.
- the client 300 and the acceleration node may be located on a same physical host, and a virtual machine on which the application program for which acceleration is requested by the client 300 is located also has same NUMA information as that of a processor to which the virtual machine belongs. Therefore, when invoking an acceleration device, the client 300 may also specify target NUMA information in the invocation request according to the requirement of the service of the application program, that is, on the basis of ensuring that cross-NUMA access to a memory corresponding to another processor does not need to be performed during service acceleration, so that the allocation unit 105 may query the acceleration device information to determine at least one candidate acceleration device whose acceleration type and algorithm type are respectively the same as the target acceleration type and the target algorithm type, whose remaining bandwidth is greater than or equal to the target acceleration bandwidth, and whose NUMA information is consistent with the target NUMA information, and determine one of the at least one candidate acceleration device as the target acceleration device.
- the remaining bandwidth is obtained by calculation according to the total bandwidth and the occupied bandwidth.
- an acceleration device whose NUMA information is consistent with the target NUMA information can be scheduled to the client, so as to ensure that a process of the service and the acceleration device are on a same NUMA, thereby improving read/write performance during storage. It should be known by persons skilled in the art that this manner is also referred to as processor affinity scheduling. Refer to the prior art for details, and the details are not described herein.
- the allocation unit 104 determines a target acceleration device from the at least one candidate acceleration device, the following cases may be included.
- the acceleration management node 100 can determine an optimal target acceleration device for the client 300, and invoke the optimal target acceleration device for the client 300, thereby implementing accurate invocation.
- the instruction unit 105 may send a configuration instruction message to the target acceleration node, to instruct the target acceleration node to respond to the invocation request.
- the configuration instruction message is used to instruct the target acceleration node to configure the target acceleration device for the client 300.
- the configuration instruction message may specifically indicate an acceleration type and an algorithm type of the target acceleration device matching the invocation request, or the configuration instruction message indicates an acceleration type, an algorithm type, and acceleration bandwidth of the target acceleration device.
- the storage unit 102 may further be configured to update previously stored acceleration device information corresponding to the target acceleration device according to the target acceleration bandwidth. It should be known that the storage unit 102 stores acceleration bandwidth of the target acceleration device before the target acceleration device is configured for the client 300.
- the acceleration bandwidth includes total bandwidth and details of unoccupied bandwidth of the target acceleration device before the target acceleration device is configured.
- the occupied bandwidth of the target acceleration device changes. The target acceleration bandwidth needs to be subtracted from the occupied bandwidth before the configuration to obtain new occupied bandwidth, and the acceleration bandwidth stored in the storage unit 102 needs to be updated by using the new occupied bandwidth.
- the acceleration bandwidth of the target acceleration device is updated for the purpose of allowing the acceleration management node 100 to subsequently allocate, in real time according to current acceleration bandwidth of the target acceleration device, an acceleration device for a new invocation request sent by the client 300.
- the storage unit 102 may further be configured to record an allocation result of the instruction unit. It should be known that the allocation result specifically indicates which acceleration device is configured for the client 300, and indicates acceleration device information and the like after the configuration, so that when subsequently finding during periodical monitoring that a service of the client 300 becomes abnormal, the acceleration management node can find an acceleration device corresponding to the abnormal service and release the acceleration device.
- the acceleration management node 100 further includes: a releasing unit 106, configured to obtain a release request sent by the client 300 for releasing the target acceleration device, and invoke the target acceleration node to release the target acceleration device.
- a releasing unit 106 configured to obtain a release request sent by the client 300 for releasing the target acceleration device, and invoke the target acceleration node to release the target acceleration device.
- the storage unit 102 is further configured to set the previously stored allocation result to invalid. Because the target acceleration device is already released, the allocation result of the target acceleration device also needs to be set to invalid, so as not to affect subsequent allocation of an acceleration device by the acceleration management node for the client.
- an embodiment of the present invention provides a schematic structural diagram of an acceleration node 200.
- the acceleration node 200 includes an agent unit 201, a driver 202, and at least one acceleration device.
- 203a and 203b are used to respectively represent two acceleration devices, and an actual quantity of acceleration devices is not limited thereto.
- the driver 202 is configured to drive the acceleration devices 203a and 203b, and the acceleration devices 203a and 203b are each configured to provide a hardware acceleration function.
- the agent unit 201 is configured to:
- the acceleration node 200 may be specifically a physical host.
- the physical host may include a memory, a processor, and at least one acceleration device (also referred to as an accelerator).
- the acceleration device may be a processor, a GPU, an FPGA, a PCI device, or the like.
- the agent unit 201 and the driver 202 may be program instructions stored in the memory.
- the processor reads the program instructions in the memory to perform corresponding functions of the agent unit 201 and the driver 202.
- the acceleration device information further includes acceleration bandwidth.
- the acceleration node reports acceleration device information of its acceleration devices to an acceleration management node, so that the acceleration management node can configure an appropriate acceleration device for a client according to the reported acceleration device information, thereby meeting a service requirement and implementing accurate invocation.
- the agent unit 201 is further configured to: receive a configuration instruction message sent by the acceleration management node 100, where if the acceleration device information includes the acceleration type and the algorithm type, the configuration instruction message indicates a target acceleration type and a target algorithm type of a target acceleration device matching an invocation request of a client, or if the acceleration device information includes the acceleration type, the algorithm type, and the acceleration bandwidth, the configuration instruction message indicates a target acceleration type, a target algorithm type, and target acceleration bandwidth of a target acceleration device matching an invocation request of a client 300; invoke, according to the configuration instruction message, the driver 202 to detect whether the target acceleration device works normally; and when the target acceleration device works normally, configure a target interface of the target acceleration device for the client 300.
- the acceleration device information further includes non-uniform memory access architecture NUMA information.
- the configuration instruction message further indicates target NUMA information matching the invocation request of the client.
- the acceleration node reports NUMA information of each acceleration device to the acceleration management node 100, so that the acceleration management node 100 allocates an appropriate target acceleration device to the client 300 according to the NUMA information, so as to ensure that the service of the client and the target acceleration device are on a same NUMA, thereby improving read/write performance during storage.
- the acceleration device specifically communicates with the client 300 by using an interface of the acceleration device.
- An acceleration device may include one or more interfaces.
- the agent unit 201 configures one of the interfaces as the target interface for the client.
- the agent unit 201 is further configured to configure the target acceleration type, the target algorithm type, and the target acceleration bandwidth as a hardware attribute of the target interface. It should be known that in the foregoing descriptions, after the target interface of the target acceleration device is configured for the client 300, the target acceleration device can provide an acceleration function for the client 300. In addition, if the acceleration bandwidth of the target acceleration device is not completely occupied, theoretically, the target acceleration device may also provide a hardware acceleration function for an application program of another client by using another interface.
- the agent unit 201 configures the target acceleration device according to the configuration instruction message sent by the acceleration management node 100, and when determining the target acceleration device, the acceleration management node uses an acceleration device whose acceleration type and algorithm type are respectively the same as the target acceleration type and the target algorithm type and whose remaining bandwidth is greater than or equal to the target acceleration bandwidth as the target acceleration device, if a target acceleration device whose unoccupied bandwidth that is far greater than the target acceleration bandwidth required by the client, an acceleration capability of the target acceleration device is wasted. Therefore, in this embodiment, after configuring the target interface for the client 300, the agent unit 201 may further configure the target acceleration type, the target algorithm type, and the target acceleration bandwidth as the hardware attribute of the target interface.
- the agent unit 201 can periodically invoke the driver 202 to query various interfaces including the target interface of the target acceleration device and obtain an acceleration device attribute of the target acceleration device in real time, so that the acceleration management node 100 can allocate the target acceleration device to another client, thereby maximizing utilization of an acceleration capability of the target acceleration device.
- the client 300 sends a release request for releasing the target acceleration device to the acceleration management node 100, so that the acceleration management node 100 invokes the agent unit 201 to release the target acceleration device. Therefore, the agent unit 201 is further configured to respond to the acceleration management node 100 and release the target acceleration device.
- agent unit 210 is further configured to set the hardware attribute of the target interface to null.
- the agent unit 210 configures the target acceleration type, the target algorithm type, and the target acceleration bandwidth as the hardware attribute of the target interface. After responding to the acceleration management node 100 and releasing the target acceleration device, correspondingly, the agent unit 201 also needs to set the hardware attribute of the target interface to null to indicate that the target interface is unoccupied, so as to prevent the agent unit 201 from obtaining incorrect acceleration device information of the target acceleration device when the proxy unit 201 periodically queries the acceleration device information.
- an embodiment of the present invention provides a schematic structural diagram of a client 300.
- the client 300 includes:
- the invocation request may further include target acceleration bandwidth required for accelerating the service.
- the invocation request may further include target NUMA information required for accelerating the service.
- the client 300 may be specifically a physical host running an application program.
- the physical host may include a memory and a processor.
- the processor reads an application program in the memory to perform a corresponding function.
- the application program may be divided into a requesting unit 301 and a sending unit 302 according to functions.
- the requesting unit 301 generates a corresponding invocation request according to an acceleration requirement of a service (that is, some functions of the application program) that needs to be offloaded to hardware for acceleration.
- the invocation request specifically includes a target acceleration type, a target algorithm type, and target acceleration bandwidth that are required for accelerating the service, or may further include target NUMA information.
- the sending unit 302 feeds back the invocation request to the acceleration management node 100 by using a communications interface between the sending unit 302 and the acceleration management node 100, so as to apply to the acceleration management node 100 for a target acceleration device matching the invocation request.
- an application program of a client may send, to an acceleration management node, a target acceleration type, a target algorithm type, and target acceleration bandwidth that correspond to an acceleration device that meets a requirement of accelerating a service of the application program, to apply to the acceleration management node for the acceleration device, so that the acceleration management node can more accurately invoke a target acceleration device required by the client.
- an acceleration type, an algorithm type, acceleration bandwidth, and target NUMA information of the target acceleration device invoked by the acceleration management node are consistent with the target acceleration type, the target algorithm type, the target acceleration bandwidth, and the target NUMA information for which the client applies, normal running of the service can be ensured.
- the requesting unit 301 is further configured to: when the acceleration of the service is completed, generate a release request for releasing the target acceleration device.
- the sending unit 302 is further configured to send the release request to the acceleration management node 100, so that the acceleration management node invokes a target acceleration node on which the target acceleration device is located to release the target acceleration device.
- the acceleration management node needs to be instructed to release a corresponding target acceleration device, so as to avoid unnecessary occupation of the target acceleration device.
- FIG. 5 to FIG. 7 Methods of the embodiments of the present invention are briefly described below with reference to FIG. 5 to FIG. 7 . It should be known that method embodiments shown in FIG. 5 to FIG. 7 are in a one-to-one correspondence with the apparatus embodiments shown in FIG. 2 to FIG. 4 . Therefore, reference can be made to each other, and descriptions are not provided again below.
- an embodiment of the present invention provides a flowchart of an acceleration management method.
- the method is applied to an acceleration management node (refer to FIG. 2 ).
- the method may include the following steps.
- S401 Separately receive acceleration device information of all acceleration devices of each of at least one acceleration node reported by the acceleration node, where each acceleration node includes at least one acceleration device, and the acceleration device information includes an acceleration type and an algorithm type.
- S402 Store the acceleration device information. It should be known that S402 is optional because a receiving unit that receives the acceleration device information may have a cache capability.
- S403 Obtain an invocation request sent by a client, where the invocation request is used to invoke an acceleration device to accelerate a service of the client, and the invocation request includes a target acceleration type and a target algorithm type.
- S404 Query the acceleration device information to determine, from all the acceleration devices of the at least one acceleration node, a target acceleration device matching the invocation request.
- S405 Instruct a target acceleration node on which the target acceleration device is located to respond to the invocation request, so that the target acceleration node configures the target acceleration device for the client.
- the acceleration management node may send a configuration instruction message to the target acceleration node where the configuration instruction message is used to instruct the target acceleration node to configure the target acceleration device for the client.
- S404 may include: querying the acceleration device information to determine, from all the acceleration devices of the at least one acceleration node, the target acceleration device whose acceleration type and algorithm type are respectively the same as the target acceleration type and the target algorithm type.
- an acceleration management node may allocate, according to a requirement of an application program of a client and an acceleration type and an algorithm type of each acceleration device, a corresponding acceleration device to the application program, thereby implementing correct invocation and ensuring normal running of the accelerated service.
- the acceleration information further includes acceleration bandwidth.
- the acceleration bandwidth of each acceleration device includes total bandwidth and occupied bandwidth.
- the invocation request further includes target acceleration bandwidth.
- S404 may specifically include: querying the acceleration device information to determine, from all the acceleration devices of the at least one acceleration node, at least one candidate acceleration device whose acceleration type and algorithm type are respectively the same as the target acceleration type and the target algorithm type and whose remaining bandwidth is greater than or equal to the target acceleration bandwidth, and determining one of the at least one candidate acceleration device as the target acceleration device, where the remaining bandwidth is obtained by calculation according to the total bandwidth and the occupied bandwidth.
- the acceleration device information received in S401 may further include NUMA information.
- the invocation request obtained in S403 may further include target NUMA information.
- S404 may specifically include: querying the acceleration device information to determine, from all the acceleration devices of the at least one acceleration node, at least one candidate acceleration device whose acceleration type and algorithm type are respectively the same as the target acceleration type and the target algorithm type, whose remaining bandwidth is greater than or equal to the target acceleration bandwidth, and whose NUMA information is consistent with the target NUMA information, and determining one of the at least one candidate acceleration device as the target acceleration device, where the remaining bandwidth is obtained by calculation according to the total bandwidth and the occupied bandwidth.
- the determining one of the at least one candidate acceleration device as the target acceleration device in S404 may include:
- S405 may include: sending a configuration instruction message to the target acceleration node, to instruct the target acceleration node to respond to the invocation request.
- the configuration instruction message is used to instruct the target acceleration node to configure the target acceleration device for the client.
- the configuration instruction message may specifically indicate an acceleration type and an algorithm type of the target acceleration device matching the invocation request, or the configuration instruction message indicates an acceleration type, an algorithm type, and acceleration bandwidth of the target acceleration device.
- the method may further include:
- Step 407a Obtain a release request sent by the client for releasing the target acceleration device, and invoke the target acceleration node to release the target acceleration device.
- S407b When detecting that the service of the client becomes abnormal, find the target acceleration device according to the recorded allocation result, and invoke the target acceleration node to release the target acceleration device.
- the acceleration management node may periodically monitor whether a service for which each client applies for acceleration runs normally. If a service becomes abnormal, the acceleration management node may find, according to the allocation result recorded in S406, a target acceleration device configured for the abnormal service, and invoke a target acceleration node on which the target acceleration device is located to release the target acceleration device, so as to prevent the target acceleration device from unnecessary operation after the service becomes abnormal.
- the method further includes: S408: Set the allocation result to invalid.
- the allocation result of the target acceleration device also needs to be set to invalid, so as not to affect subsequent allocation of an acceleration device by the acceleration management node for the client.
- an embodiment of the present invention further provides a flowchart of an acceleration device configuration method.
- the method is applied to an acceleration node (refer to FIG. 3 ).
- the acceleration node includes a driver and at least one acceleration device.
- the method includes the following steps.
- S501 Invoke the driver to separately query the at least one acceleration device to obtain acceleration device information of each acceleration device.
- the acceleration device information includes an acceleration type and an algorithm type.
- S502 Report the acceleration device information to an acceleration management node.
- an acceleration node reports acceleration device information of its acceleration devices to an acceleration management node, so that the acceleration management node can configure an appropriate acceleration device for a client according to the reported acceleration device information, thereby meeting a service requirement and implementing accurate invocation.
- the acceleration device information may further include acceleration bandwidth.
- the method further includes the following steps.
- S503 Receive a configuration instruction message sent by the acceleration management node. If the acceleration device information includes the acceleration type and the algorithm type, the configuration instruction message indicates a target acceleration type and a target algorithm type of a target acceleration device matching the invocation request of the client, or if the acceleration device information includes the acceleration type, the algorithm type, and the acceleration bandwidth, the configuration instruction message indicates a target acceleration type, a target algorithm type, and target acceleration bandwidth of a target acceleration device matching the invocation request of the client.
- S504 Invoke, according to the configuration instruction message, the driver to detect whether the target acceleration device works normally.
- the application program of client can invoke the target interface and run a service to be accelerated of the application program.
- acceleration devices may be grouped according to processors to which the acceleration devices respectively belong. Therefore, in step S501, the acceleration device information may further include NUMA information.
- the NUMA information may indicate a grouping status of each acceleration device.
- the configuration instruction message received in step S503 may further indicate target NUMA information matching the invocation request of the client.
- S506 Configure the target acceleration type, the target algorithm type, and the target acceleration bandwidth as a hardware attribute of the target interface.
- the acceleration node can subsequently query the remaining acceleration bandwidth of the target acceleration device, so that the acceleration management node can allocate the target acceleration device to another client, thereby maximizing utilization of an acceleration capability of the target acceleration device.
- S507 Respond to the acceleration management node and release the target acceleration device.
- S508 Set the hardware attribute of the target interface to null.
- an embodiment of the present invention further provides a flowchart of a method of applying for an acceleration device.
- the method is applied to a client.
- the method includes the following steps.
- S601 Generate an invocation request according to an acceleration requirement of a service.
- the invocation request includes a target acceleration type and a target algorithm type that are required for accelerating the service.
- S602 Send the invocation request to an acceleration management node to request to invoke a target acceleration device matching the invocation request to accelerate the service.
- the acceleration management node invokes a target acceleration device according to the target acceleration type and the target algorithm type, so that an acceleration type and an algorithm type of the target acceleration device match the acceleration requirement of the service of the client, thereby implementing accurate invocation.
- step S601 if acceleration device information reported by an acceleration node to the acceleration management node includes acceleration bandwidth, the invocation request generated by the client further includes target acceleration bandwidth required for accelerating the service.
- an application program of a client may send, to an acceleration management node, a target acceleration type, a target algorithm type, and target acceleration bandwidth that correspond to an acceleration device that meets a requirement of accelerating a service of the application program, to apply to the acceleration management node for the acceleration device, so that the acceleration management node can more accurately invoke a target acceleration device required by the client, and ensure normal running of the service.
- acceleration devices in the acceleration node may be grouped according to NUMA information. Therefore, the invocation request may further include target NUMA information required by the service, so that the service and the target acceleration device required by the service are configured on a same NUMA.
- the client and the acceleration node may actually be a same physical host, and the application program and the agent unit in the acceleration node are different processes or different software modules. Therefore, configuring the service and target acceleration device on a same NUMA can ensure that the service and the target acceleration device read a memory in the same NUMA, thereby improving read performance.
- S604 Send the release request to the acceleration management node.
- the client may instruct, by performing steps S603 and S604, the acceleration management node to release the corresponding target acceleration device, so as to avoid unnecessary occupation of the target acceleration device.
- An embodiment of the present invention further provides an acceleration management system.
- the acceleration management system includes an acceleration management node 100 and at least one acceleration node 200.
- the acceleration management node 100 refer to the acceleration management node shown in FIG. 2 and the corresponding embodiment.
- the acceleration node 200 refer to the acceleration node shown in FIG. 3 and the corresponding embodiment. Details are not described herein again.
- the acceleration management system provided by this embodiment of the present invention can accurately invoke, according to a service requirement of an application program of a client, an appropriate acceleration device to accelerate the service of the application program, and ensure normal running of the service.
- the program may be stored in a computer-readable storage medium.
- the foregoing storage medium includes: any medium that can store program code, such as a ROM, a RAM, a magnetic disk, or an optical disc.
Abstract
Description
- This application claims priority to Chinese Patent Application No.
201510628762.0 - The present invention relates to the field of virtualization technologies, and in particular, to an acceleration management node, an acceleration node, a client, and a method.
- To shorten an execution time of an application program and improve the running efficiency, usually some services (or functions) in the program may be allocated to a hardware acceleration device for execution. Because the hardware acceleration device runs fast, the execution time of the application program can be shortened. Common acceleration services include encryption and decryption, compression, decompression, audio and video encoding and decoding, and the like. Hardware acceleration devices include a processor that provides particular instructions and other peripheral component interconnect (peripheral component interconnect, PCI) devices that can provide an acceleration function, such as a graphics processing unit (graphics processing unit, GPU) and a field programmable gate array (field programmable gate array, FPGA).
- At present, network function virtualization (Network Function Virtualization, NFV) is proposed for the purpose of implementing some network functions in general-purpose high-performance servers, switches, and storage devices by using a virtualization technology. In a network function virtualization (Network Function Virtualization, NFV) evolution scenario, a network device based on special-purpose hardware may be deployed on a general-purpose server by using a virtualization technology, to evolve from a conventional combination form of "embedded software + special-purpose hardware" to a combination form of "software + general-purpose hardware". That is, to implement hardware generalization, a network function (Network Function, NF) program needs to be separated from conventional special-purpose hardware to form a virtualization network function (Virtualization Network Function, VNF) program, so that the conventional special-purpose hardware becomes general-purpose NFV hardware.
- However, after NFV hardware is generalized, how to accurately schedule the NFV hardware according to a service requirement of an application program when the NFV hardware is invoked becomes a problem to be resolved urgently.
- Embodiments of the present invention provide an acceleration management node, an acceleration node, a client, and a method, applied to a virtualization scenario, so that an acceleration device can be accurately invoked according to a service requirement of a client.
- According to a first aspect, an embodiment of the present invention provides an acceleration management node, where the acceleration management node includes: a receiving unit, configured to separately receive acceleration device information of all acceleration devices of each of at least one acceleration node reported by the acceleration node, where each acceleration node includes at least one acceleration device, and the acceleration device information includes an acceleration type and an algorithm type; an obtaining unit, configured to obtain an invocation request sent by a client, where the invocation request is used to invoke an acceleration device to accelerate a service of the client, and the invocation request includes a target acceleration type and a target algorithm type; an allocation unit, configured to query the acceleration device information to determine, from all the acceleration devices of the at least one acceleration node, a target acceleration device matching the invocation request; and an instruction unit, configured to instruct a target acceleration node on which the target acceleration device is located to respond to the invocation request. The acceleration management node invokes an acceleration device according to acceleration device information in each acceleration node and may allocate, according to a requirement of an application program of a client and an acceleration type and an algorithm type of each acceleration device, a corresponding acceleration device to the application program, so as to meet a service requirement.
- In a first possible implementation manner of the first aspect, the allocation unit is configured to query the acceleration device information to determine, from all the acceleration devices of the at least one acceleration node, the target acceleration device whose acceleration type and algorithm type are respectively the same as the target acceleration type and the target algorithm type. When invoking an acceleration device according to acceleration device information in each acceleration node, the acceleration management node performs invocation according to an acceleration type, an algorithm type, and acceleration bandwidth of each acceleration device, so as to ensure that bandwidth of the acceleration device can meet a service requirement, thereby implementing accurate invocation.
- With reference to the first aspect or the first possible implementation manner of the first aspect, in a second possible implementation manner, the acceleration information further includes acceleration bandwidth, and the acceleration bandwidth includes total bandwidth and occupied bandwidth; the invocation request further includes target acceleration bandwidth; and the allocation unit is specifically configured to query the acceleration device information to determine at least one candidate acceleration device whose acceleration type and algorithm type are respectively the same as the target acceleration type and the target algorithm type and whose remaining bandwidth is greater than or equal to the target acceleration bandwidth, and determine one of the at least one candidate acceleration device as the target acceleration device, where the remaining bandwidth is obtained by calculation according to the total bandwidth and the occupied bandwidth.
- With reference to the second possible implementation manner of the first aspect, in a third possible implementation manner, the acceleration device information further includes non-uniform memory access architecture NUMA information; the invocation request further includes target NUMA information; and the allocation unit is specifically configured to query the acceleration device information to determine at least one candidate acceleration device whose acceleration type and algorithm type are respectively the same as the target acceleration type and the target algorithm type, whose remaining bandwidth is greater than or equal to the target acceleration bandwidth, and whose NUMA information is consistent with the target NUMA information, and determine one of the at least one candidate acceleration device as the target acceleration device, where the remaining bandwidth is obtained by calculation according to the total bandwidth and the occupied bandwidth.
- With reference to either of the second possible implementation manner and the third possible implementation manner of the first aspect, in a fourth possible implementation manner, the allocation unit is specifically configured to: when there is one candidate acceleration device, determine the candidate acceleration device as the target acceleration device.
- With reference to any one of the second to fourth possible implementation manners of the first aspect, in a fifth possible implementation manner, the allocation unit is specifically configured to: when there is a plurality of candidate acceleration devices, determine a first acceleration device having maximum remaining bandwidth from the plurality of candidate acceleration devices according to the acceleration bandwidth, and if there is one first acceleration device, determine the first acceleration device as the target acceleration device.
- With reference to the fifth possible implementation manner of the first aspect, in a sixth possible implementation manner, the allocation unit is specifically configured to: when there is a plurality of first acceleration devices having the maximum remaining bandwidth, determine a second acceleration device having a maximum VF quantity from the plurality of first acceleration devices according to the VF quantity, and if there is one second acceleration device, use the second acceleration device as the target acceleration device.
- With reference to the sixth possible implementation manner of the first aspect, in a seventh possible implementation manner, the allocation unit is specifically configured to: when there is a plurality of second acceleration devices having the maximum VF quantity, use a second acceleration device first found as the target acceleration device according to a time sequence of querying the acceleration device information.
- With reference to any one of the first aspect or the first to seventh possible implementation manners of the first aspect, in an eighth possible implementation manner, the instruction unit is specifically configured to send configuration instruction information to the target acceleration node, to instruct the target acceleration node to respond to the invocation request, where the configuration instruction message indicates an acceleration type and an algorithm type of the target acceleration device matching the invocation request, or the configuration instruction message indicates an acceleration type, an algorithm type, and acceleration bandwidth of the target acceleration device.
- With reference to any one of the first aspect or the first to eighth possible implementation manners of the first aspect, in a ninth possible implementation manner, the acceleration management node further includes a storage unit, configured to store the acceleration device information.
- With reference to any one of the second to ninth possible implementation manners of the first aspect, in a tenth possible implementation manner, the storage unit is further configured to: update previously stored acceleration device information corresponding to the target acceleration device according to the target acceleration bandwidth; and record an allocation result of the instruction unit.
- With reference to any one of the first aspect or the first to tenth possible implementation manners of the first aspect, in an eleventh possible implementation manner, the acceleration management node further includes: a releasing unit, configured to obtain a release request sent by the client for releasing the target acceleration device, and invoke the target acceleration node to release the target acceleration device.
- With reference to the eleventh possible implementation manner of the first aspect, in a twelfth possible implementation manner, the releasing unit is configured to: when detecting that the service of the client becomes abnormal, find the target acceleration device according to the allocation result recorded by the storage unit, and invoke the target acceleration node to release the target acceleration device.
- With reference to any one of the ninth to twelfth possible implementation manners of the first aspect, in a thirteenth possible implementation manner, the storage unit is further configured to set the allocation result to invalid.
- According to a second aspect, an embodiment of the present invention provides an acceleration node, including: an agent unit, a driver, and at least one acceleration device, where the driver is configured to drive the at least one acceleration device, the at least one acceleration device is configured to provide a hardware acceleration function, and the agent unit is configured to: invoke the driver to separately query the at least one acceleration device to obtain acceleration device information of each acceleration device, where the acceleration device information includes an acceleration type and an algorithm type; and report the acceleration device information to an acceleration management node. The acceleration node reports acceleration device information of its acceleration devices to the acceleration management node, so that the acceleration management node can configure an appropriate acceleration device for a client according to the reported acceleration device information, thereby meeting a service requirement and implementing accurate invocation.
- In a first possible implementation manner of the second aspect, the agent unit is further configured to: acceleration bandwidth.
- With reference to the second aspect or the first possible implementation manner of the second aspect, in a second possible implementation manner, the agent unit is further configured to: receive a configuration instruction message sent by the acceleration management node, where the configuration instruction message indicates a target acceleration type and a target algorithm type of a target acceleration device matching an invocation request of a client, or the configuration instruction message indicates a target acceleration type, a target algorithm type, and target acceleration bandwidth of a target acceleration device matching an invocation request of a client; invoke, according to the configuration instruction message, the driver to detect whether the target acceleration device works normally; and when the target acceleration device works normally, configure a target interface of the target acceleration device for the client.
- With reference to any one of the second aspect or the first to second possible implementation manners of the second aspect, in a third possible implementation manner, the acceleration device information further includes non-uniform memory access architecture NUMA information.
- With reference to the third possible implementation manner of the second aspect, in a fourth possible implementation manner, the configuration instruction message further indicates target NUMA information matching the invocation request of the client.
- With reference to any one of the second to fourth possible implementation manners of the second aspect, in a fifth possible implementation manner, the agent unit is further configured to configure the target acceleration type, the target algorithm type, and the target acceleration bandwidth as a hardware attribute of the target interface.
- With reference to any one of the second to fifth possible implementation manners of the second aspect, in a sixth possible implementation manner, the agent unit is further configured to respond to the acceleration management node and release the target acceleration device.
- With reference to the sixth possible implementation manner of the second aspect, in a seventh possible implementation manner, the agent unit is further configured to set the hardware attribute of the target interface to null.
- According to a third aspect, an embodiment of the present invention provides a client, including: a requesting unit, configured to generate an invocation request according to an acceleration requirement of a service, where the invocation request includes a target acceleration type and a target algorithm type that are required for accelerating the service; and a sending unit, configured to send the invocation request to an acceleration management node to request to invoke a target acceleration device matching the invocation request to accelerate the service. An application program of the client may send, to the acceleration management node, a target acceleration type and a target algorithm type that correspond to an acceleration device that meets a requirement of accelerating a service of the application program, to apply to the acceleration management node for the acceleration device, so that the acceleration management node can more accurately invoke a target acceleration device required by the client.
- In a first possible implementation manner of the third aspect, the invocation request further includes target acceleration bandwidth required for accelerating the service.
- With reference to the third aspect or the first possible implementation manner of the third aspect, in a second possible implementation manner, the invocation request further includes target NUMA information required by the service.
- With reference to any one of the third aspect or the first to second possible implementation manners of the third aspect, in a third possible implementation manner, the requesting unit is further configured to: when the client needs to release the acceleration device, generate a release request for releasing the target acceleration device; and the sending unit is further configured to send the release request to the acceleration management node.
- According to a fourth aspect, an embodiment of the present invention provides an acceleration management method, including: separately receiving acceleration device information of all acceleration devices of each of at least one acceleration node reported by the acceleration node, where each acceleration node includes at least one acceleration device, and the acceleration device information includes an acceleration type and an algorithm type; obtaining an invocation request sent by a client, where the invocation request is used to invoke an acceleration device to accelerate a service of the client, and the invocation request includes a target acceleration type and a target algorithm type; querying the acceleration device information to determine, from all the acceleration devices of the at least one acceleration node, a target acceleration device matching the invocation request; and instructing a target acceleration node on which the target acceleration device is located to respond to the invocation request. In this method, by obtaining acceleration device information of each acceleration node and invoking an acceleration device, an acceleration management node may allocate, according to a requirement of an application program of a client and an acceleration type and an algorithm type of each acceleration device, a corresponding acceleration device to the application program, thereby ensuring normal running of the accelerated service and implementing accurate invocation.
- In a first possible implementation manner of the fourth aspect, the step of querying the acceleration device information to determine, from all the acceleration devices of the at least one acceleration node, a target acceleration device matching the invocation request includes: querying the acceleration device information to determine the target acceleration device whose acceleration type and algorithm type are respectively the same as the target acceleration type and the target algorithm type.
- With reference to the fourth aspect or the first possible implementation manner of the fourth aspect, in a second possible implementation manner, the acceleration information further includes acceleration bandwidth, and the acceleration bandwidth includes total bandwidth and occupied bandwidth; the invocation request further includes target acceleration bandwidth; and the step of querying the acceleration device information to determine, from all the acceleration devices of the at least one acceleration node, a target acceleration device matching the invocation request includes: querying the acceleration device information to determine at least one candidate acceleration device whose acceleration type and algorithm type are respectively the same as the target acceleration type and the target algorithm type and whose remaining bandwidth is greater than or equal to the target acceleration bandwidth, and determine one of the at least one candidate acceleration device as the target acceleration device, where the remaining bandwidth is obtained by calculation according to the total bandwidth and the occupied bandwidth.
- With reference to the second possible implementation manner of the fourth aspect, in a third possible implementation manner, the acceleration device information further includes non-uniform memory access architecture NUMA information; the invocation request further includes target NUMA information; and the step of querying the acceleration device information to determine, from all the acceleration devices of the at least one acceleration node, a target acceleration device matching the invocation request specifically includes: querying the acceleration device information to determine at least one candidate acceleration device whose acceleration type and algorithm type are respectively the same as the target acceleration type and the target algorithm type, whose remaining bandwidth is greater than or equal to the target acceleration bandwidth, and whose NUMA information is consistent with the target NUMA information, and determine one of the at least one candidate acceleration device as the target acceleration device, where the remaining bandwidth is obtained by calculation according to the total bandwidth and the occupied bandwidth.
- With reference to either of the second possible implementation manner and the third possible implementation manner of the fourth aspect, in a fourth possible implementation manner, when there is one candidate acceleration device, the candidate acceleration device is determined as the target acceleration device; or when there is a plurality of candidate acceleration devices, a first acceleration device having maximum remaining bandwidth is determined from the plurality of candidate acceleration devices according to the acceleration bandwidth, and if there is one first acceleration device, the first acceleration device is determined as the target acceleration device bandwidth.
- With reference to the fourth possible implementation manner of the fourth aspect, in a fifth possible implementation manner, the acceleration device information further includes a virtual function VF quantity; and when there is a plurality of first acceleration devices having the maximum remaining bandwidth, a second acceleration device having a maximum VF quantity is determined from the plurality of first acceleration devices, and if there is one second acceleration device, the second acceleration device is used as the target acceleration device.
- With reference to the fifth possible implementation manner of the fourth aspect, in a sixth possible implementation manner, when there is a plurality of second acceleration devices having the maximum VF quantity, a second acceleration device first found is used as the target acceleration device according to a time sequence of querying the acceleration device information.
- With reference to any one of the fourth aspect or the first to sixth possible implementation manners of the fourth aspect, in a seventh possible implementation manner, the method further includes: storing the acceleration device information.
- With reference to any one of the second to seventh possible implementation manners of the fourth aspect, in an eighth possible implementation manner, the method further includes: updating previously stored acceleration device information corresponding to the target acceleration device according to the target acceleration bandwidth; and recording an allocation result.
- With reference to any one of the fourth aspect or the first to eighth possible implementation manners of the fourth aspect, in a ninth possible implementation manner, the method further includes: obtaining a release request sent by the client for releasing the target acceleration device, and invoking the target acceleration node to release the target acceleration device.
- With reference to the ninth possible implementation manner of the fourth aspect, in a tenth possible implementation manner, the method further includes: when detecting that the service of the client becomes abnormal, finding the target acceleration device according to the recorded allocation result, and invoking the target acceleration node to release the target acceleration device.
- With reference to any one of the eighth to tenth possible implementation manners of the fourth aspect, in an eleventh possible implementation manner, the method further includes: setting the allocation result to invalid.
- According to a fifth aspect, an embodiment of the present invention provides an acceleration device configuration method, applied to an acceleration node, where the acceleration node includes a driver and at least one acceleration device; and the method includes: invoking the driver to separately query the at least one acceleration device to obtain acceleration device information of each acceleration device, where the acceleration device information includes an acceleration type and an algorithm type; and reporting the acceleration device information to an acceleration management node. The acceleration node reports acceleration device information of its acceleration devices to the acceleration management node, so that the acceleration management node can configure an appropriate acceleration device for a client according to the reported acceleration device information, thereby meeting a service requirement and implementing accurate invocation.
- In a first possible implementation manner of the fifth aspect, the acceleration device information further includes acceleration bandwidth.
- With reference to the fifth aspect or the first possible implementation manner of the fifth aspect, in a second possible implementation manner, the method further includes: receiving a configuration instruction message sent by the acceleration management node, where the configuration instruction message indicates a target acceleration type and a target algorithm type of a target acceleration device matching an invocation request of a client, or the configuration instruction message indicates a target acceleration type, a target algorithm type, and target acceleration bandwidth of a target acceleration device matching an invocation request of a client; invoking, according to the configuration instruction message, the driver to detect whether the target acceleration device works normally; and when the target acceleration device works normally, configuring a target interface of the target acceleration device for the client.
- With reference to the fifth aspect or the second possible implementation manner of the fifth aspect, in a third possible implementation manner, the acceleration device information further includes non-uniform memory access architecture NUMA information.
- With reference to the third possible implementation manner of the fifth aspect, in a fourth possible implementation manner, the configuration instruction message further indicates target NUMA information matching the invocation request of the client.
- With reference to either of the second possible implementation manner and the fourth possible implementation manner of the fifth aspect, in a fifth possible implementation manner, the method further includes: configuring the target acceleration type, the target algorithm type, and the target acceleration bandwidth as a hardware attribute of the target interface.
- With reference to any one of the second to fifth possible implementation manners of the fifth aspect, in a sixth possible implementation manner, the method further includes: responding to the acceleration management node and releasing the target acceleration device.
- With reference to the sixth possible implementation manner of the fifth aspect, in a seventh possible implementation manner, the method further includes: setting the hardware attribute of the target interface to null.
- According to a sixth aspect, an embodiment of the present invention provides a method of applying for an acceleration device, including: generating an invocation request according to an acceleration requirement of a service, where the invocation request includes a target acceleration type and a target algorithm type that are required for accelerating the service; and sending the invocation request to an acceleration management node to request to invoke a target acceleration device matching the invocation request to accelerate the service. An application program of a client may send, to the acceleration management node, a target acceleration type and a target algorithm type that correspond to an acceleration device that meets a requirement of accelerating a service of the application program, to apply to the acceleration management node for the acceleration device, so that the acceleration management node can more accurately invoke a target acceleration device required by the client, and ensure normal running of the service.
- In a first possible implementation manner of the sixth aspect, the invocation request further includes target acceleration bandwidth required for accelerating the service.
- With reference to the sixth aspect or the first possible implementation manner of the sixth aspect, in a second possible implementation manner, the invocation request further includes target NUMA information required by the service.
- With reference to any one of the sixth aspect or the first to second possible implementation manners of the sixth aspect, in a third possible implementation manner, the method further includes: when the acceleration of the service is completed, generating a release request for releasing the target acceleration device; and sending the release request to the acceleration management node.
- According to a seventh aspect, an embodiment of the present invention provides an acceleration management system, including: the acceleration management node according to any one of the first aspect or the first to thirteenth possible implementation manners of the first aspect and the acceleration node according to any one of the second aspect or the first to seventh possible implementation manners. The acceleration management system can accurately invoke, according to a service requirement of an application program of a client, an appropriate acceleration device to accelerate a service of the application program and ensure normal running of the service.
- To describe the technical solutions in the embodiments of the present invention or in the prior art more clearly, the following briefly describes the accompanying drawings required for describing the embodiments. Apparently, the accompanying drawings in the following description show merely some embodiments of the present invention, and persons of ordinary skill in the art may still derive other drawings from these accompanying drawings without creative efforts.
-
FIG. 1 is a schematic diagram of an application scenario of a virtualization technology; -
FIG. 2 is a schematic structural diagram of an acceleration management node according to an embodiment of the present invention; -
FIG. 3 is a schematic structural diagram of an acceleration node according to an embodiment of the present invention; -
FIG. 4 is a schematic structural diagram of a client according to an embodiment of the present invention; -
FIG. 5 is a flowchart of an acceleration management method according to an embodiment of the present invention; -
FIG. 6 is a flowchart of an acceleration device configuration method according to an embodiment of the present invention; and -
FIG. 7 is a flowchart of a method of applying for an acceleration device according to an embodiment of the present invention. - The following clearly describes the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Apparently, the described embodiments are merely some but not all of the embodiments of the present invention. All other embodiments obtained by persons of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.
- First, it should be noted that the technical solutions of the present invention may be applied to various virtualization scenarios. Virtualization refers to virtualizing one computer into a plurality of logical computers by using a virtualization technology. For example, as shown in
FIG. 1 , a computation management module (not shown) in a client 300 (also referred to as a computer or a physical host) may create at least one virtual machine according to a user requirement. For example,FIG. 1 shows three virtual machines in theclient 300, namely, avirtual machine 300a, avirtual machine 300b, and avirtual machine 300c. The virtual machines may run on different operating systems. Each virtual machine may be considered as a logical computer. Because all application programs of the virtual machines can run in mutually independent spaces without affecting each other, the working efficiency of theclient 300 is significantly improved. InFIG. 1 , anacceleration node 200 may include general-purpose computer hardware, that is, an acceleration device such as a CPU or a GPU, and each virtual machine in theclient 300 may invoke the acceleration device in theacceleration node 200 by using anacceleration management node 100. Network function virtualization (Network Function Virtualization, NFV) uses general-purpose hardware such as x86 and a virtualization technology to implement software processing of a lot of functions, so as to reduce network device costs. NFV may be considered as an actual application of the virtualization technology. In addition, the virtualization technology may also be applied in scenarios such as a public cloud, a private cloud, an enterprise cloud, and cloud acceleration. Therefore, the solutions of the embodiments of the present invention are not limited to an NFV scenario, and the protection scope of the present invention should not be limited thereto. - To better describe the technical solutions of the present invention, the technical solutions of the present invention are described in detail below with reference to
FIG. 2 to FIG. 4 . -
FIG. 2 is a schematic structural diagram of anacceleration management node 100 according to an embodiment of the present invention. Theacceleration management node 100 includes: a receivingunit 101, an obtainingunit 103, anallocation unit 104, and aninstruction unit 105. - The receiving
unit 101 is configured to separately receive acceleration device information of all acceleration devices of each of at least one acceleration node reported by the acceleration node. Each acceleration node includes at least one acceleration device. The acceleration device information includes an acceleration type and an algorithm type. The acceleration type indicates a type of an acceleration function supported by each acceleration device. For example, common acceleration functions may include encryption, decryption, compression, decompression, audio and video encoding and decoding, and the like. The algorithm type indicates an algorithm used by each acceleration device when the acceleration device implements an acceleration function supported by the acceleration device. - To better describe the technical solutions of the present invention, as shown in
FIG. 2 , in some virtualization scenarios, for example,FIG. 2 shows three acceleration nodes, namely, anacceleration node 200a, anacceleration node 200b, and anacceleration node 200c. An actual quantity of acceleration nodes may be set according to a network requirement. No limitation is imposed herein. Each acceleration node may include at least one hardware acceleration device such as a CPU, a GPU, or a PCI device. Each acceleration device has its own acceleration type, algorithm type, and the like. That is, each acceleration device corresponds to one piece of acceleration device information. Different acceleration devices may correspond to same or different acceleration device information. When an acceleration node includes two acceleration devices, namely, a CPU and a GPU, the acceleration node may separately report acceleration device information of the CPU and acceleration device information of the GPU to theacceleration management node 100 by using the receivingunit 101. In addition, the receivingunit 101 may be a reporting interface and may specifically include a software interface. For example, theacceleration node 200a may invoke the reporting interface in a Remote Procedure Call Protocol (Remote Procedure Call Protocol, RPC) manner, to report the acceleration device information of all the acceleration devices in theacceleration node 200a to theacceleration management node 100. Theother acceleration nodes acceleration node 200a, and details are not described herein again. - It should be known that the
acceleration management node 100 in this embodiment of the present invention may be a management program running on a physical host. The physical host may usually include a processor, a memory, and an input/output (I/O) interface. The management program is stored in the memory. The processor can read and run the management program in the memory. Further, the receivingunit 101 may be a software I/O interface, and the acceleration node may use various communications tools (for example, a communications tool Rabbit MQ) between software I/O interfaces to remotely invoke the software I/O interface for communication. It should be known by persons skilled in the art that communication may also be performed between software I/O interfaces by using various other message queues. No specific limitation is imposed herein. - The obtaining
unit 103 is configured to obtain an invocation request sent by aclient 300. The invocation request is used to invoke an acceleration device to accelerate a service of theclient 300, and the invocation request includes a target acceleration type and a target algorithm type. - It should be noted that the
client 300 may be specifically a physical host that runs an application program. When a service (or function) of the application program needs to be accelerated by using an acceleration device to shorten an execution time of the application program, the application program of theclient 300 may notify, by using the obtainingunit 103, theacceleration management node 100 of information, such as a target acceleration type and a target algorithm type, of an acceleration device required for accelerating the service, so as to apply to theacceleration management node 100 for a hardware resource (that is, an acceleration device) for acceleration. The obtainingunit 103 may also be an application programming interface (application programming interface, API). The application program of theclient 300 may communicate with the obtainingunit 103 by invoking an application programming interface. It should be known that each type of service of the application program requires a target acceleration type and a target algorithm type that comply with its service specification. - The
allocation unit 104 is configured to query the acceleration device information to determine, from all the acceleration devices of the at least one acceleration node, a target acceleration device matching the invocation request. - It should be noted that after obtaining the invocation request sent by the
client 300, theallocation unit 104 may search acceleration device information of all acceleration nodes that is stored in astorage unit 102, for a target acceleration device that meets the target acceleration type and the target algorithm type required by the invocation request. - The
instruction unit 105 is configured to instruct a target acceleration node on which the target acceleration device is located to respond to the invocation request and configure the target acceleration device for theclient 300. - Optionally, the
acceleration management node 100 further includes: thestorage unit 102, configured to the acceleration device information. - It should be noted that the
storage unit 102 may store the obtained acceleration device information in a local memory of theacceleration management node 100 or in a network memory connected to theacceleration management node 100 through a network. No limitation is imposed herein. In addition, thestorage unit 102 may store the acceleration device information in a list form, a database form, or another storage form well-known to persons skilled in the art. - For example, it is assumed that after the
client 300 requests theacceleration management node 100 to invoke an acceleration device to accelerate the service, theacceleration management node 100 determines, by querying the acceleration device information obtained from theacceleration nodes client 300 is located on theacceleration node 200c. In this case, theinstruction unit 105 may invoke a configuration interface of theacceleration node 200c in an RPC manner to allocate the target acceleration type and the target algorithm type to theacceleration node 200c, so that theacceleration node 200c configures the corresponding target acceleration device for theclient 300, thereby providing a hardware acceleration function for the service of the application program of theclient 300. - In this embodiment, the acceleration management node invokes an acceleration device according to acceleration device information in each acceleration node and may allocate, according to a requirement of an application program of a client and an acceleration type and an algorithm type of each acceleration device, a corresponding acceleration device to the application program, so as to meet a service requirement. It should be known by persons skilled in the art that acceleration refers to allocating some services of an application program to a hardware acceleration device for operation. Because efficiency of logical operations of the hardware acceleration device is higher than that of a software algorithm, an operation time can be saved, thereby achieving acceleration. However, in the prior art, when an acceleration device is invoked, a specific acceleration type and algorithm type supported by the acceleration device are not considered. Using acceleration of encryption and decryption as an example, only an encryption and decryption type, such as IPSec, can be perceived in the prior art, and invocation is performed according to the encryption and decryption type IPSec. However, the encryption and decryption type IPSec further includes three sub-types, namely, 3DES (Triple Data Encryption Algorithm, Triple Data Encryption Algorithm), DH (Diffie-Hellman Algorithm, D-H algorithm), and AES (Advanced Encryption Standard, Advanced Encryption Standard). In the prior art, an acceleration device of the IPSec-DH type may be invoked for a service that requires an acceleration device of the IPSec-3DES type, resulting in that an invocation result cannot meet the requirement of the service. Compared with the prior art, the technical solutions of the present invention can ensure accuracy of an invocation result, so that an attribute of an acceleration device invoked by the
client 300 can meet a requirement of an application program. - In this embodiment, the
allocation unit 104 may be specifically configured to query the acceleration device information to determine, from all the acceleration devices of the at least one acceleration node, the target acceleration device whose acceleration type and algorithm type are respectively the same as the target acceleration type and the target algorithm type. - In this embodiment, further, the acceleration information obtained by the receiving
unit 101 may further include acceleration bandwidth. The acceleration bandwidth may include total bandwidth of an acceleration device and details of occupied bandwidth of the acceleration device at a current moment. The total bandwidth of the acceleration device refers to maximum acceleration bandwidth that can be provided by the acceleration device in a zero-load state. Correspondingly, the invocation request obtained by the obtainingunit 103 may further include target acceleration bandwidth. The target acceleration bandwidth indicates bandwidth required by a service for which acceleration is requested by theclient 300. - In this embodiment, because the acceleration bandwidth of each acceleration device includes the total bandwidth and the occupied bandwidth, the
allocation unit 104 may first query the acceleration device information to obtain at least one candidate acceleration device whose acceleration type and algorithm type are respectively the same as the target acceleration type and the target algorithm type and whose remaining bandwidth is greater than or equal to the target acceleration bandwidth, that is, to obtain at least one candidate acceleration device matching the invocation request; and determine one of the at least one candidate acceleration device as the target acceleration device. The remaining bandwidth is obtained by calculation according to the total bandwidth and the occupied bandwidth. - As can be seen, the
acceleration management node 100 may allocate a corresponding acceleration device to the application program according to acceleration bandwidth of each acceleration device, so as to ensure that the allocated acceleration device can provide sufficient acceleration bandwidth for the application program, thereby implementing accurate invocation. However, in the prior art, the acceleration bandwidth of an acceleration device is not considered, and when remaining bandwidth of the acceleration device does not meet a service requirement, a prolonged acceleration time or an acceleration failure may also be caused, failing to achieve acceleration. - Further, when the
acceleration node 200 uses a multi-processor architecture, theacceleration node 200 may provide a separate memory for each processor by using a non-uniform memory access architecture (Non Uniform Memory Access Architecture, NUMA), thereby avoiding performance loss caused by access of multiple processors to a same memory. Therefore, processors in theacceleration node 200 may be grouped according to NUMA information, and the processors in different groups have different NUMA information. In addition, the acceleration devices in theacceleration node 200 usually belong to different processors. Therefore, each acceleration device has same NUMA information as that of the processor to which the acceleration device belongs. Correspondingly, the acceleration device information obtained by the receivingunit 101 may further include NUMA information. Moreover, theclient 300 and the acceleration node may be located on a same physical host, and a virtual machine on which the application program for which acceleration is requested by theclient 300 is located also has same NUMA information as that of a processor to which the virtual machine belongs. Therefore, when invoking an acceleration device, theclient 300 may also specify target NUMA information in the invocation request according to the requirement of the service of the application program, that is, on the basis of ensuring that cross-NUMA access to a memory corresponding to another processor does not need to be performed during service acceleration, so that theallocation unit 105 may query the acceleration device information to determine at least one candidate acceleration device whose acceleration type and algorithm type are respectively the same as the target acceleration type and the target algorithm type, whose remaining bandwidth is greater than or equal to the target acceleration bandwidth, and whose NUMA information is consistent with the target NUMA information, and determine one of the at least one candidate acceleration device as the target acceleration device. The remaining bandwidth is obtained by calculation according to the total bandwidth and the occupied bandwidth. In such a manner, an acceleration device whose NUMA information is consistent with the target NUMA information can be scheduled to the client, so as to ensure that a process of the service and the acceleration device are on a same NUMA, thereby improving read/write performance during storage. It should be known by persons skilled in the art that this manner is also referred to as processor affinity scheduling. Refer to the prior art for details, and the details are not described herein. - Because there may be one or more candidate acceleration devices, when the
allocation unit 104 determines a target acceleration device from the at least one candidate acceleration device, the following cases may be included. - (1) When there is one candidate acceleration device, the candidate acceleration device is determined as the target acceleration device.
- (2) When there is a plurality of candidate acceleration devices, the allocation unit determines a first acceleration device having maximum remaining bandwidth from the plurality of candidate acceleration devices according to the acceleration bandwidth. If there is one first acceleration device, the first acceleration device is determined as the target acceleration device. The remaining bandwidth of the candidate acceleration device is obtained by calculation according to the total bandwidth and the occupied bandwidth. In addition, in this embodiment, if a third acceleration device whose remaining bandwidth is approximately equal to the target acceleration bandwidth exists in the plurality of candidate acceleration devices, the third acceleration device may also be determined as the target acceleration device. It should be known by persons skilled in the art that being approximately equal refers to that a slight difference in value is allowed.
- (3) When there is a plurality of candidate acceleration devices, and there is a plurality of first acceleration devices having the maximum remaining bandwidth in the plurality of the candidate acceleration devices, the acceleration device information reported by the
acceleration node 200 to theacceleration management node 100 further includes a virtual function (Virtual Function, VF) quantity. The allocation unit determines a second acceleration device having a maximum VF quantity from the plurality of first acceleration devices according to the VF quantity, and if there is one second acceleration device, uses the second acceleration device as the target acceleration device. - (4) When there is a plurality of candidate acceleration devices, and there is a plurality of first acceleration devices having the maximum remaining bandwidth in the plurality of the candidate acceleration devices, if there is a plurality of second acceleration devices having the maximum VF quantity, the allocation unit uses a second acceleration device first found as the target acceleration device according to a time sequence of querying the acceleration device information.
- In the foregoing manner, the
acceleration management node 100 can determine an optimal target acceleration device for theclient 300, and invoke the optimal target acceleration device for theclient 300, thereby implementing accurate invocation. - In this embodiment, for example, specifically, the
instruction unit 105 may send a configuration instruction message to the target acceleration node, to instruct the target acceleration node to respond to the invocation request. The configuration instruction message is used to instruct the target acceleration node to configure the target acceleration device for theclient 300. The configuration instruction message may specifically indicate an acceleration type and an algorithm type of the target acceleration device matching the invocation request, or the configuration instruction message indicates an acceleration type, an algorithm type, and acceleration bandwidth of the target acceleration device. - Further, in this embodiment, the
storage unit 102 may further be configured to update previously stored acceleration device information corresponding to the target acceleration device according to the target acceleration bandwidth. It should be known that thestorage unit 102 stores acceleration bandwidth of the target acceleration device before the target acceleration device is configured for theclient 300. The acceleration bandwidth includes total bandwidth and details of unoccupied bandwidth of the target acceleration device before the target acceleration device is configured. After the target acceleration device is configured for theclient 300 to provide hardware acceleration for the service of theclient 300, correspondingly, the occupied bandwidth of the target acceleration device changes. The target acceleration bandwidth needs to be subtracted from the occupied bandwidth before the configuration to obtain new occupied bandwidth, and the acceleration bandwidth stored in thestorage unit 102 needs to be updated by using the new occupied bandwidth. The acceleration bandwidth of the target acceleration device is updated for the purpose of allowing theacceleration management node 100 to subsequently allocate, in real time according to current acceleration bandwidth of the target acceleration device, an acceleration device for a new invocation request sent by theclient 300. In addition, thestorage unit 102 may further be configured to record an allocation result of the instruction unit. It should be known that the allocation result specifically indicates which acceleration device is configured for theclient 300, and indicates acceleration device information and the like after the configuration, so that when subsequently finding during periodical monitoring that a service of theclient 300 becomes abnormal, the acceleration management node can find an acceleration device corresponding to the abnormal service and release the acceleration device. - Further, in this embodiment, the
acceleration management node 100 further includes:
a releasing unit 106, configured to obtain a release request sent by theclient 300 for releasing the target acceleration device, and invoke the target acceleration node to release the target acceleration device. Still using theacceleration node 200c as an example, because the target acceleration device located on theacceleration node 200c is configured for theclient 300, when the application program in theclient 300 needs to release the acceleration device, theclient 300 may instruct, by using a release request, theacceleration management node 100 to release the target acceleration device. - Still further, after the target acceleration device is released, the
storage unit 102 is further configured to set the previously stored allocation result to invalid. Because the target acceleration device is already released, the allocation result of the target acceleration device also needs to be set to invalid, so as not to affect subsequent allocation of an acceleration device by the acceleration management node for the client. - As shown in
FIG. 3 , an embodiment of the present invention provides a schematic structural diagram of anacceleration node 200. Theacceleration node 200 includes anagent unit 201, adriver 202, and at least one acceleration device. For example, 203a and 203b are used to respectively represent two acceleration devices, and an actual quantity of acceleration devices is not limited thereto. It should be known by persons skilled in the art that thedriver 202 is configured to drive theacceleration devices acceleration devices agent unit 201 is configured to: - invoke the
driver 202 to separately query the at least one acceleration device to obtain acceleration device information of each acceleration device, where the acceleration device information includes an acceleration type and an algorithm type, and for example, theagent unit 201 may periodically invoke thedriver 202 to query each interface on theacceleration devices acceleration devices - report the obtained acceleration device information to the
acceleration management node 100. Specifically, after obtaining the acceleration device information by query, theagent unit 201 may invoke the receivingunit 101 in theacceleration management node 100 in an RPC manner to report the acceleration device information to theacceleration management node 100. - It should be noted that, in this embodiment, in this embodiment, the
acceleration node 200 may be specifically a physical host. The physical host may include a memory, a processor, and at least one acceleration device (also referred to as an accelerator). The acceleration device may be a processor, a GPU, an FPGA, a PCI device, or the like. Theagent unit 201 and thedriver 202 may be program instructions stored in the memory. The processor reads the program instructions in the memory to perform corresponding functions of theagent unit 201 and thedriver 202. - In this embodiment, the acceleration device information further includes acceleration bandwidth.
- In this embodiment, the acceleration node reports acceleration device information of its acceleration devices to an acceleration management node, so that the acceleration management node can configure an appropriate acceleration device for a client according to the reported acceleration device information, thereby meeting a service requirement and implementing accurate invocation.
- Further, in this embodiment, the
agent unit 201 is further configured to:
receive a configuration instruction message sent by theacceleration management node 100, where if the acceleration device information includes the acceleration type and the algorithm type, the configuration instruction message indicates a target acceleration type and a target algorithm type of a target acceleration device matching an invocation request of a client, or if the acceleration device information includes the acceleration type, the algorithm type, and the acceleration bandwidth, the configuration instruction message indicates a target acceleration type, a target algorithm type, and target acceleration bandwidth of a target acceleration device matching an invocation request of aclient 300; invoke, according to the configuration instruction message, thedriver 202 to detect whether the target acceleration device works normally; and when the target acceleration device works normally, configure a target interface of the target acceleration device for theclient 300. - Further, in this embodiment, when the
acceleration node 200 and theclient 300 are a same physical host, and the physical host uses a multi-core architecture, the acceleration device information further includes non-uniform memory access architecture NUMA information. Correspondingly, the configuration instruction message further indicates target NUMA information matching the invocation request of the client. - The acceleration node reports NUMA information of each acceleration device to the
acceleration management node 100, so that theacceleration management node 100 allocates an appropriate target acceleration device to theclient 300 according to the NUMA information, so as to ensure that the service of the client and the target acceleration device are on a same NUMA, thereby improving read/write performance during storage. - It should be noted that when providing an acceleration function for the
client 300, the acceleration device specifically communicates with theclient 300 by using an interface of the acceleration device. An acceleration device may include one or more interfaces. During configuration, theagent unit 201 configures one of the interfaces as the target interface for the client. - Still further, the
agent unit 201 is further configured to configure the target acceleration type, the target algorithm type, and the target acceleration bandwidth as a hardware attribute of the target interface. It should be known that in the foregoing descriptions, after the target interface of the target acceleration device is configured for theclient 300, the target acceleration device can provide an acceleration function for theclient 300. In addition, if the acceleration bandwidth of the target acceleration device is not completely occupied, theoretically, the target acceleration device may also provide a hardware acceleration function for an application program of another client by using another interface. However, because theagent unit 201 configures the target acceleration device according to the configuration instruction message sent by theacceleration management node 100, and when determining the target acceleration device, the acceleration management node uses an acceleration device whose acceleration type and algorithm type are respectively the same as the target acceleration type and the target algorithm type and whose remaining bandwidth is greater than or equal to the target acceleration bandwidth as the target acceleration device, if a target acceleration device whose unoccupied bandwidth that is far greater than the target acceleration bandwidth required by the client, an acceleration capability of the target acceleration device is wasted. Therefore, in this embodiment, after configuring the target interface for theclient 300, theagent unit 201 may further configure the target acceleration type, the target algorithm type, and the target acceleration bandwidth as the hardware attribute of the target interface. In this way, subsequently theagent unit 201 can periodically invoke thedriver 202 to query various interfaces including the target interface of the target acceleration device and obtain an acceleration device attribute of the target acceleration device in real time, so that theacceleration management node 100 can allocate the target acceleration device to another client, thereby maximizing utilization of an acceleration capability of the target acceleration device. - In this embodiment, further, after the acceleration of the service of the
client 300 is completed, theclient 300 sends a release request for releasing the target acceleration device to theacceleration management node 100, so that theacceleration management node 100 invokes theagent unit 201 to release the target acceleration device. Therefore, theagent unit 201 is further configured to respond to theacceleration management node 100 and release the target acceleration device. - Still further, the agent unit 210 is further configured to set the hardware attribute of the target interface to null.
- In the foregoing descriptions, to maximize the utilization of the acceleration capability of the target acceleration device, the agent unit 210 configures the target acceleration type, the target algorithm type, and the target acceleration bandwidth as the hardware attribute of the target interface. After responding to the
acceleration management node 100 and releasing the target acceleration device, correspondingly, theagent unit 201 also needs to set the hardware attribute of the target interface to null to indicate that the target interface is unoccupied, so as to prevent theagent unit 201 from obtaining incorrect acceleration device information of the target acceleration device when theproxy unit 201 periodically queries the acceleration device information. - As shown in
FIG. 4 , an embodiment of the present invention provides a schematic structural diagram of aclient 300. Theclient 300 includes: - a requesting
unit 301, configured to generate an invocation request according to an acceleration requirement of a service, where the invocation request includes a target acceleration type and a target algorithm type that are required for accelerating the service; and - a sending
unit 302, configured to send the invocation request to an acceleration management node to request to invoke a target acceleration device matching the invocation request to accelerate the service. - Further, the invocation request may further include target acceleration bandwidth required for accelerating the service.
- Still further, the invocation request may further include target NUMA information required for accelerating the service.
- It should be noted that, in this embodiment, the
client 300 may be specifically a physical host running an application program. The physical host may include a memory and a processor. The processor reads an application program in the memory to perform a corresponding function. The application program may be divided into a requestingunit 301 and a sendingunit 302 according to functions. The requestingunit 301 generates a corresponding invocation request according to an acceleration requirement of a service (that is, some functions of the application program) that needs to be offloaded to hardware for acceleration. The invocation request specifically includes a target acceleration type, a target algorithm type, and target acceleration bandwidth that are required for accelerating the service, or may further include target NUMA information. The sendingunit 302 feeds back the invocation request to theacceleration management node 100 by using a communications interface between the sendingunit 302 and theacceleration management node 100, so as to apply to theacceleration management node 100 for a target acceleration device matching the invocation request. - By means of the solution of this embodiment of the present invention, an application program of a client may send, to an acceleration management node, a target acceleration type, a target algorithm type, and target acceleration bandwidth that correspond to an acceleration device that meets a requirement of accelerating a service of the application program, to apply to the acceleration management node for the acceleration device, so that the acceleration management node can more accurately invoke a target acceleration device required by the client. In addition, because an acceleration type, an algorithm type, acceleration bandwidth, and target NUMA information of the target acceleration device invoked by the acceleration management node are consistent with the target acceleration type, the target algorithm type, the target acceleration bandwidth, and the target NUMA information for which the client applies, normal running of the service can be ensured.
- Further, the requesting
unit 301 is further configured to: when the acceleration of the service is completed, generate a release request for releasing the target acceleration device. - The sending
unit 302 is further configured to send the release request to theacceleration management node 100, so that the acceleration management node invokes a target acceleration node on which the target acceleration device is located to release the target acceleration device. - That is, after the acceleration of the service of the application program of the client is completed, the acceleration management node needs to be instructed to release a corresponding target acceleration device, so as to avoid unnecessary occupation of the target acceleration device.
- Methods of the embodiments of the present invention are briefly described below with reference to
FIG. 5 to FIG. 7 . It should be known that method embodiments shown inFIG. 5 to FIG. 7 are in a one-to-one correspondence with the apparatus embodiments shown inFIG. 2 to FIG. 4 . Therefore, reference can be made to each other, and descriptions are not provided again below. - As shown in
FIG. 5 , an embodiment of the present invention provides a flowchart of an acceleration management method. The method is applied to an acceleration management node (refer toFIG. 2 ). The method may include the following steps. - S401: Separately receive acceleration device information of all acceleration devices of each of at least one acceleration node reported by the acceleration node, where each acceleration node includes at least one acceleration device, and the acceleration device information includes an acceleration type and an algorithm type.
- S402: Store the acceleration device information. It should be known that S402 is optional because a receiving unit that receives the acceleration device information may have a cache capability.
- S403: Obtain an invocation request sent by a client, where the invocation request is used to invoke an acceleration device to accelerate a service of the client, and the invocation request includes a target acceleration type and a target algorithm type.
- S404: Query the acceleration device information to determine, from all the acceleration devices of the at least one acceleration node, a target acceleration device matching the invocation request.
- S405: Instruct a target acceleration node on which the target acceleration device is located to respond to the invocation request, so that the target acceleration node configures the target acceleration device for the client. For example, the acceleration management node may send a configuration instruction message to the target acceleration node where the configuration instruction message is used to instruct the target acceleration node to configure the target acceleration device for the client.
- More specifically, S404 may include: querying the acceleration device information to determine, from all the acceleration devices of the at least one acceleration node, the target acceleration device whose acceleration type and algorithm type are respectively the same as the target acceleration type and the target algorithm type.
- In this embodiment, by obtaining acceleration device information of each acceleration node and invoking an acceleration device, an acceleration management node may allocate, according to a requirement of an application program of a client and an acceleration type and an algorithm type of each acceleration device, a corresponding acceleration device to the application program, thereby implementing correct invocation and ensuring normal running of the accelerated service.
- Further, the acceleration information further includes acceleration bandwidth. The acceleration bandwidth of each acceleration device includes total bandwidth and occupied bandwidth. Correspondingly, the invocation request further includes target acceleration bandwidth.
- S404 may specifically include:
querying the acceleration device information to determine, from all the acceleration devices of the at least one acceleration node, at least one candidate acceleration device whose acceleration type and algorithm type are respectively the same as the target acceleration type and the target algorithm type and whose remaining bandwidth is greater than or equal to the target acceleration bandwidth, and determining one of the at least one candidate acceleration device as the target acceleration device, where the remaining bandwidth is obtained by calculation according to the total bandwidth and the occupied bandwidth. - Still further, the acceleration device information received in S401 may further include NUMA information. The invocation request obtained in S403 may further include target NUMA information. S404 may specifically include:
querying the acceleration device information to determine, from all the acceleration devices of the at least one acceleration node, at least one candidate acceleration device whose acceleration type and algorithm type are respectively the same as the target acceleration type and the target algorithm type, whose remaining bandwidth is greater than or equal to the target acceleration bandwidth, and whose NUMA information is consistent with the target NUMA information, and determining one of the at least one candidate acceleration device as the target acceleration device, where the remaining bandwidth is obtained by calculation according to the total bandwidth and the occupied bandwidth. - Further, the determining one of the at least one candidate acceleration device as the target acceleration device in S404 may include:
- when there is one candidate acceleration device, determining the candidate acceleration device as the target acceleration device;
- when there is a plurality of candidate acceleration devices, determining a first acceleration device having maximum remaining bandwidth from the plurality of candidate acceleration devices according to the acceleration bandwidth, and if there is one first acceleration device, determining the first acceleration device as the target acceleration device bandwidth, where the remaining bandwidth of the candidate acceleration device may be obtained by calculation according to the total bandwidth and the occupied bandwidth;
- when there is a plurality of candidate acceleration devices, and there is a plurality of first acceleration devices having the maximum remaining bandwidth in the plurality of the candidate acceleration devices, determining a second acceleration device having a maximum VF quantity from the plurality of first acceleration devices, and if there is one second acceleration device, using the second acceleration device as the target acceleration device, where in S401, the acceleration device information received by the acceleration management node includes a VF quantity that is supported by each acceleration device in the acceleration node and that is reported by the acceleration node; or
- when there is a plurality of candidate acceleration devices, and there is a plurality of first acceleration devices having the maximum remaining bandwidth in the plurality of the candidate acceleration devices, if there is a plurality of second acceleration devices having the maximum VF quantity, using a second acceleration device first found as the target acceleration device according to a time sequence of querying the acceleration device information.
- In this embodiment, for example, S405 may include: sending a configuration instruction message to the target acceleration node, to instruct the target acceleration node to respond to the invocation request. The configuration instruction message is used to instruct the target acceleration node to configure the target acceleration device for the client. The configuration instruction message may specifically indicate an acceleration type and an algorithm type of the target acceleration device matching the invocation request, or the configuration instruction message indicates an acceleration type, an algorithm type, and acceleration bandwidth of the target acceleration device.
- In this embodiment, further, the method may further include:
- S406: Update previously stored acceleration device information corresponding to the target acceleration device according to the target acceleration bandwidth; and record an allocation result, where the allocation result indicates which acceleration device is allocated as the target acceleration device to the client by the acceleration management node according to the invocation request. It should be noted that if the acceleration device information only includes the acceleration type and the algorithm type, only the allocation result of the acceleration management node is recorded in S406.
- S406 may be performed after S405 or may be performed at the same time as S405. No limitation is imposed herein.
- Further, the method further includes:
Step 407a: Obtain a release request sent by the client for releasing the target acceleration device, and invoke the target acceleration node to release the target acceleration device. - Alternatively, S407b: When detecting that the service of the client becomes abnormal, find the target acceleration device according to the recorded allocation result, and invoke the target acceleration node to release the target acceleration device. The acceleration management node may periodically monitor whether a service for which each client applies for acceleration runs normally. If a service becomes abnormal, the acceleration management node may find, according to the allocation result recorded in S406, a target acceleration device configured for the abnormal service, and invoke a target acceleration node on which the target acceleration device is located to release the target acceleration device, so as to prevent the target acceleration device from unnecessary operation after the service becomes abnormal.
- Still further, after S407a or S407b, the method further includes:
S408: Set the allocation result to invalid. - Because the target acceleration device is already released, the allocation result of the target acceleration device also needs to be set to invalid, so as not to affect subsequent allocation of an acceleration device by the acceleration management node for the client.
- As shown in
FIG. 6 , an embodiment of the present invention further provides a flowchart of an acceleration device configuration method. The method is applied to an acceleration node (refer toFIG. 3 ). The acceleration node includes a driver and at least one acceleration device. The method includes the following steps. - S501: Invoke the driver to separately query the at least one acceleration device to obtain acceleration device information of each acceleration device. The acceleration device information includes an acceleration type and an algorithm type.
- S502: Report the acceleration device information to an acceleration management node.
- In this embodiment, an acceleration node reports acceleration device information of its acceleration devices to an acceleration management node, so that the acceleration management node can configure an appropriate acceleration device for a client according to the reported acceleration device information, thereby meeting a service requirement and implementing accurate invocation.
- Further, the acceleration device information may further include acceleration bandwidth.
- Further, in this embodiment, the method further includes the following steps.
- S503: Receive a configuration instruction message sent by the acceleration management node. If the acceleration device information includes the acceleration type and the algorithm type, the configuration instruction message indicates a target acceleration type and a target algorithm type of a target acceleration device matching the invocation request of the client, or if the acceleration device information includes the acceleration type, the algorithm type, and the acceleration bandwidth, the configuration instruction message indicates a target acceleration type, a target algorithm type, and target acceleration bandwidth of a target acceleration device matching the invocation request of the client.
- S504: Invoke, according to the configuration instruction message, the driver to detect whether the target acceleration device works normally.
- S505: When the target acceleration device works normally, configure a target interface of the target acceleration device for the client.
- After the target interface of the target acceleration device is configured for the client, the application program of client can invoke the target interface and run a service to be accelerated of the application program.
- When the acceleration node uses a multi-processor architecture, acceleration devices may be grouped according to processors to which the acceleration devices respectively belong. Therefore, in step S501, the acceleration device information may further include NUMA information. The NUMA information may indicate a grouping status of each acceleration device.
- Correspondingly, the configuration instruction message received in step S503 may further indicate target NUMA information matching the invocation request of the client.
- S506: Configure the target acceleration type, the target algorithm type, and the target acceleration bandwidth as a hardware attribute of the target interface. By means of the configuration of the hardware attribute of the target interface, the acceleration node can subsequently query the remaining acceleration bandwidth of the target acceleration device, so that the acceleration management node can allocate the target acceleration device to another client, thereby maximizing utilization of an acceleration capability of the target acceleration device.
- S507: Respond to the acceleration management node and release the target acceleration device. S508: Set the hardware attribute of the target interface to null.
- As shown in
FIG. 7 , an embodiment of the present invention further provides a flowchart of a method of applying for an acceleration device. The method is applied to a client. The method includes the following steps. - S601: Generate an invocation request according to an acceleration requirement of a service. The invocation request includes a target acceleration type and a target algorithm type that are required for accelerating the service.
- S602: Send the invocation request to an acceleration management node to request to invoke a target acceleration device matching the invocation request to accelerate the service. The acceleration management node invokes a target acceleration device according to the target acceleration type and the target algorithm type, so that an acceleration type and an algorithm type of the target acceleration device match the acceleration requirement of the service of the client, thereby implementing accurate invocation.
- Further, in step S601, if acceleration device information reported by an acceleration node to the acceleration management node includes acceleration bandwidth, the invocation request generated by the client further includes target acceleration bandwidth required for accelerating the service.
- In this embodiment, an application program of a client may send, to an acceleration management node, a target acceleration type, a target algorithm type, and target acceleration bandwidth that correspond to an acceleration device that meets a requirement of accelerating a service of the application program, to apply to the acceleration management node for the acceleration device, so that the acceleration management node can more accurately invoke a target acceleration device required by the client, and ensure normal running of the service.
- In this embodiment, when the acceleration node uses a multi-processor architecture, acceleration devices in the acceleration node may be grouped according to NUMA information. Therefore, the invocation request may further include target NUMA information required by the service, so that the service and the target acceleration device required by the service are configured on a same NUMA. It should be noted that in this embodiment, the client and the acceleration node may actually be a same physical host, and the application program and the agent unit in the acceleration node are different processes or different software modules. Therefore, configuring the service and target acceleration device on a same NUMA can ensure that the service and the target acceleration device read a memory in the same NUMA, thereby improving read performance.
- S603: When the acceleration of the service is completed, generate a release request for releasing the target acceleration device.
- S604: Send the release request to the acceleration management node.
- After the acceleration of the service of the application program of the client is completed, the client may instruct, by performing steps S603 and S604, the acceleration management node to release the corresponding target acceleration device, so as to avoid unnecessary occupation of the target acceleration device.
- An embodiment of the present invention further provides an acceleration management system. Referring to
FIG. 1 , the acceleration management system includes anacceleration management node 100 and at least oneacceleration node 200. For theacceleration management node 100, refer to the acceleration management node shown inFIG. 2 and the corresponding embodiment. For theacceleration node 200, refer to the acceleration node shown inFIG. 3 and the corresponding embodiment. Details are not described herein again. The acceleration management system provided by this embodiment of the present invention can accurately invoke, according to a service requirement of an application program of a client, an appropriate acceleration device to accelerate the service of the application program, and ensure normal running of the service. - Persons of ordinary skill in the art may understand that all or some of the steps of the method embodiments may be implemented by a program instructing relevant hardware. The program may be stored in a computer-readable storage medium. When the program runs, the steps of the method embodiments are performed. The foregoing storage medium includes: any medium that can store program code, such as a ROM, a RAM, a magnetic disk, or an optical disc.
- Finally, it should be noted that the foregoing embodiments are merely intended for describing the technical solutions of the present invention, but not for limiting the present invention. Although the present invention is described in detail with reference to the foregoing embodiments, persons of ordinary skill in the art should understand that they may still make modifications to the technical solutions described in the foregoing embodiments or make equivalent replacements to some or all technical features thereof, without departing from the scope of the technical solutions of the embodiments of the present invention.
Claims (51)
- An acceleration management node, comprising:a receiving unit, configured to separately receive acceleration device information of all acceleration devices of each of at least one acceleration node reported by the acceleration node, wherein each acceleration node comprises at least one acceleration device, and the acceleration device information comprises an acceleration type and an algorithm type;an obtaining unit, configured to obtain an invocation request sent by a client, wherein the invocation request is used to invoke an acceleration device to accelerate a service of the client, and the invocation request comprises a target acceleration type and a target algorithm type;an allocation unit, configured to query the acceleration device information to determine, from all the acceleration devices of the at least one acceleration node, a target acceleration device matching the invocation request; andan instruction unit, configured to instruct a target acceleration node on which the target acceleration device is located to respond to the invocation request.
- The acceleration management node according to claim 1, wherein the allocation unit is configured to query the acceleration device information to determine, from all the acceleration devices of the at least one acceleration node, the target acceleration device whose acceleration type and algorithm type are respectively the same as the target acceleration type and the target algorithm type.
- The acceleration management node according to claim 1 or 2, wherein the acceleration information further comprises acceleration bandwidth, and the acceleration bandwidth comprises total bandwidth and occupied bandwidth; the invocation request further comprises target acceleration bandwidth; and
the allocation unit is specifically configured to query the acceleration device information to determine at least one candidate acceleration device whose acceleration type and algorithm type are respectively the same as the target acceleration type and the target algorithm type and whose remaining bandwidth is greater than or equal to the target acceleration bandwidth, and determine one of the at least one candidate acceleration device as the target acceleration device, wherein the remaining bandwidth is obtained by calculation according to the total bandwidth and the occupied bandwidth. - The acceleration management node according to claim 3, wherein the acceleration device information further comprises non-uniform memory access architecture NUMA information; the invocation request further comprises target NUMA information; and
the allocation unit is specifically configured to query the acceleration device information to determine at least one candidate acceleration device whose acceleration type and algorithm type are respectively the same as the target acceleration type and the target algorithm type, whose remaining bandwidth is greater than or equal to the target acceleration bandwidth, and whose NUMA information is consistent with the target NUMA information, and determine one of the at least one candidate acceleration device as the target acceleration device, wherein the remaining bandwidth is obtained by calculation according to the total bandwidth and the occupied bandwidth. - The acceleration management node according to claim 3 or 4, wherein
the allocation unit is specifically configured to: when there is one candidate acceleration device, determine the candidate acceleration device as the target acceleration device. - The acceleration management node according to any one of claims 3 to 5, wherein the allocation unit is specifically configured to: when there is a plurality of candidate acceleration devices, determine a first acceleration device having maximum remaining bandwidth from the plurality of candidate acceleration devices according to the acceleration bandwidth, and if there is one first acceleration device, determine the first acceleration device as the target acceleration device.
- The acceleration management node according to claim 6, wherein
the allocation unit is specifically configured to: when there is a plurality of first acceleration devices having the maximum remaining bandwidth, determine a second acceleration device having a maximum VF quantity from the plurality of first acceleration devices according to the VF quantity, and if there is one second acceleration device, use the second acceleration device as the target acceleration device. - The acceleration management node according to claim 7, wherein the allocation unit is specifically configured to: when there is a plurality of second acceleration devices having the maximum VF quantity, use a second acceleration device first found as the target acceleration device according to a time sequence of querying the acceleration device information.
- The acceleration management node according to any one of claims 1 to 8, wherein the instruction unit is specifically configured to send configuration instruction information to the target acceleration node, to instruct the target acceleration node to respond to the invocation request, wherein the configuration instruction message indicates an acceleration type and an algorithm type of the target acceleration device matching the invocation request, or the configuration instruction message indicates an acceleration type, an algorithm type, and acceleration bandwidth of the target acceleration device.
- The acceleration management node according to any one of claims 1 to 9, wherein the acceleration management node further comprises a storage unit, configured to store the acceleration device information.
- The acceleration management node according to any one of claims 3 to 10, wherein the storage unit is further configured to: update previously stored acceleration device information corresponding to the target acceleration device according to the target acceleration bandwidth; and record an allocation result of the instruction unit.
- The acceleration management node according to any one of claims 1 to 11, wherein the acceleration management node further comprises:
a releasing unit, configured to obtain a release request sent by the client for releasing the target acceleration device, and invoke the target acceleration node to release the target acceleration device. - The acceleration management node according to claim 12, wherein the releasing unit is configured to: when detecting that the service of the client becomes abnormal, find the target acceleration device according to the allocation result recorded by the storage unit, and invoke the target acceleration node to release the target acceleration device.
- The acceleration management node according to any one of claims 10 to 13, wherein the storage unit is further configured to set the allocation result to invalid.
- An acceleration node, comprising: an agent unit, a driver, and at least one acceleration device, wherein the driver is configured to drive the at least one acceleration device, the at least one acceleration device is configured to provide a hardware acceleration function, and the agent unit is configured to:invoke the driver to separately query the at least one acceleration device to obtain acceleration device information of each acceleration device, wherein the acceleration device information comprises an acceleration type and an algorithm type; andreport the acceleration device information to an acceleration management node.
- The acceleration node according to claim 15, wherein the acceleration device information further comprises acceleration bandwidth.
- The acceleration node according to claim 15 or 16, wherein the agent unit is further configured to:receive a configuration instruction message sent by the acceleration management node, wherein the configuration instruction message indicates a target acceleration type and a target algorithm type of a target acceleration device matching an invocation request of a client, or the configuration instruction message indicates a target acceleration type, a target algorithm type, and target acceleration bandwidth of a target acceleration device matching an invocation request of a client;invoke, according to the configuration instruction message, the driver to detect whether the target acceleration device works normally; andwhen the target acceleration device works normally, configure a target interface of the target acceleration device for the client.
- The acceleration node according to any one of claims 15 to 17, wherein the acceleration device information further comprises non-uniform memory access architecture NUMA information.
- The acceleration node according to claim 18, wherein the configuration instruction message further indicates target NUMA information matching the invocation request of the client.
- The acceleration node according to any one of claims 17 to 19, wherein the agent unit is further configured to configure the target acceleration type, the target algorithm type, and the target acceleration bandwidth as a hardware attribute of the target interface.
- The acceleration node according to any one of claims 17 to 20, wherein the agent unit is further configured to respond to the acceleration management node and release the target acceleration device.
- The acceleration node according to claim 21, wherein the agent unit is further configured to set the hardware attribute of the target interface to null.
- A client, comprising:a requesting unit, configured to generate an invocation request according to an acceleration requirement of a service, wherein the invocation request comprises a target acceleration type and a target algorithm type that are required for accelerating the service; anda sending unit, configured to send the invocation request to an acceleration management node to request to invoke a target acceleration device matching the invocation request to accelerate the service.
- The client according to claim 23, wherein the invocation request further comprises target acceleration bandwidth required for accelerating the service.
- The client according to claim 23 or 24, wherein the invocation request further comprises target NUMA information required by the service.
- The client according to any one of claims 23 to 25, wherein the requesting unit is further configured to: when the client needs to release the acceleration device, generate a release request for releasing the target acceleration device; and
the sending unit is further configured to send the release request to the acceleration management node. - An acceleration management method, comprising:separately receiving acceleration device information of all acceleration devices of each of at least one acceleration node reported by the acceleration node, wherein each acceleration node comprises at least one acceleration device, and the acceleration device information comprises an acceleration type and an algorithm type;obtaining an invocation request sent by a client, wherein the invocation request is used to invoke an acceleration device to accelerate a service of the client, and the invocation request comprises a target acceleration type and a target algorithm type;querying the acceleration device information to determine, from all the acceleration devices of the at least one acceleration node, a target acceleration device matching the invocation request; andinstructing a target acceleration node on which the target acceleration device is located to respond to the invocation request.
- The method according to claim 27, wherein the step of querying the acceleration device information to determine, from all the acceleration devices of the at least one acceleration node, a target acceleration device matching the invocation request comprises:
querying the acceleration device information to determine the target acceleration device whose acceleration type and algorithm type are respectively the same as the target acceleration type and the target algorithm type. - The method according to claim 27 or 28, wherein the acceleration information further comprises acceleration bandwidth, and the acceleration bandwidth comprises total bandwidth and occupied bandwidth; the invocation request further comprises target acceleration bandwidth; and
the step of querying the acceleration device information to determine, from all the acceleration devices of the at least one acceleration node, a target acceleration device matching the invocation request comprises:
querying the acceleration device information to determine at least one candidate acceleration device whose acceleration type and algorithm type are respectively the same as the target acceleration type and the target algorithm type and whose remaining bandwidth is greater than or equal to the target acceleration bandwidth, and determine one of the at least one candidate acceleration device as the target acceleration device, wherein the remaining bandwidth is obtained by calculation according to the total bandwidth and the occupied bandwidth. - The method according to claim 29, wherein the acceleration device information further comprises non-uniform memory access architecture NUMA information; the invocation request further comprises target NUMA information; and
the step of querying the acceleration device information to determine, from all the acceleration devices of the at least one acceleration node, a target acceleration device matching the invocation request specifically comprises:
querying the acceleration device information to determine at least one candidate acceleration device whose acceleration type and algorithm type are respectively the same as the target acceleration type and the target algorithm type, whose remaining bandwidth is greater than or equal to the target acceleration bandwidth, and whose NUMA information is consistent with the target NUMA information, and determine one of the at least one candidate acceleration device as the target acceleration device, wherein the remaining bandwidth is obtained by calculation according to the total bandwidth and the occupied bandwidth. - The method according to claim 29 or 30, wherein
when there is one candidate acceleration device, the candidate acceleration device is determined as the target acceleration device; or
when there is a plurality of candidate acceleration devices, a first acceleration device having maximum remaining bandwidth is determined from the plurality of candidate acceleration devices according to the acceleration bandwidth, and if there is one first acceleration device, the first acceleration device is determined as the target acceleration device bandwidth. - The method according to claim 31, wherein the acceleration device information further comprises a virtual function VF quantity; and
when there is a plurality of first acceleration devices having the maximum remaining bandwidth, a second acceleration device having a maximum VF quantity is determined from the plurality of first acceleration devices, and if there is one second acceleration device, the second acceleration device is used as the target acceleration device. - The method according to claim 32, wherein when there is a plurality of second acceleration devices having the maximum VF quantity, a second acceleration device first found is used as the target acceleration device according to a time sequence of querying the acceleration device information.
- The method according to any one of claims 27 to 33, further comprising:
storing the acceleration device information. - The method according to any one of claims 29 to 34, further comprising:
updating previously stored acceleration device information corresponding to the target acceleration device according to the target acceleration bandwidth; and recording an allocation result. - The method according to any one of claims 27 to 35, further comprising:
obtaining a release request sent by the client for releasing the target acceleration device, and invoking the target acceleration node to release the target acceleration device. - The method according to claim 36, further comprising: when detecting that the service of the client becomes abnormal, finding the target acceleration device according to the recorded allocation result, and invoking the target acceleration node to release the target acceleration device.
- The method according to any one of claims 35 to 37, further comprising: setting the allocation result to invalid.
- An acceleration device configuration method, applied to an acceleration node, wherein the acceleration node comprises a driver and at least one acceleration device; and the method comprises:invoking the driver to separately query the at least one acceleration device to obtain acceleration device information of each acceleration device, wherein the acceleration device information comprises an acceleration type and an algorithm type; andreporting the acceleration device information to an acceleration management node.
- The method according to claim 39, wherein the acceleration device information further comprises acceleration bandwidth.
- The method according to claim 39 or 40, further comprising:receiving a configuration instruction message sent by the acceleration management node, wherein the configuration instruction message indicates a target acceleration type and a target algorithm type of a target acceleration device matching an invocation request of a client, or the configuration instruction message indicates a target acceleration type, a target algorithm type, and target acceleration bandwidth of a target acceleration device matching an invocation request of a client;invoking, according to the configuration instruction message, the driver to detect whether the target acceleration device works normally; andwhen the target acceleration device works normally, configuring a target interface of the target acceleration device for the client.
- The method according to claim 39 or 41, wherein the acceleration device information further comprises non-uniform memory access architecture NUMA information.
- The method according to claim 42, wherein the configuration instruction message further indicates target NUMA information matching the invocation request of the client.
- The method according to any one of claims 41 to 43, further comprising:
configuring the target acceleration type, the target algorithm type, and the target acceleration bandwidth as a hardware attribute of the target interface. - The method according to any one of claims 41 to 44, further comprising:
responding to the acceleration management node and releasing the target acceleration device. - The method according to claim 45, further comprising:
setting the hardware attribute of the target interface to null. - A method of applying for an acceleration device, comprising:generating an invocation request according to an acceleration requirement of a service, wherein the invocation request comprises a target acceleration type and a target algorithm type that are required for accelerating the service; andsending the invocation request to an acceleration management node to request to invoke a target acceleration device matching the invocation request to accelerate the service.
- The method according to claim 47, wherein the invocation request further comprises target acceleration bandwidth required for accelerating the service.
- The method according to claim 47 or 48, wherein the invocation request further comprises target NUMA information required by the service.
- The method according to any one of claims 47 to 49, further comprising:when the acceleration of the service is completed, generating a release request for releasing the target acceleration device; andsending the release request to the acceleration management node.
- An acceleration management system, comprising: the acceleration management node according to any one of claims 1 to 14 and the acceleration node according to any one of claims 15 to 22.
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
EP20174899.3A EP3767912B1 (en) | 2015-09-28 | 2016-09-26 | Acceleration management methods |
EP21217676.2A EP4040761A3 (en) | 2015-09-28 | 2016-09-26 | Acceleration management node, computer program product and non-transitory computer-readable medium |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510628762.0A CN105357258B (en) | 2015-09-28 | 2015-09-28 | Acceleration management node, acceleration node, client and method |
PCT/CN2016/100137 WO2017054691A1 (en) | 2015-09-28 | 2016-09-26 | Acceleration management node, acceleration node, client, and method |
Related Child Applications (3)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
EP20174899.3A Division-Into EP3767912B1 (en) | 2015-09-28 | 2016-09-26 | Acceleration management methods |
EP20174899.3A Division EP3767912B1 (en) | 2015-09-28 | 2016-09-26 | Acceleration management methods |
EP21217676.2A Division EP4040761A3 (en) | 2015-09-28 | 2016-09-26 | Acceleration management node, computer program product and non-transitory computer-readable medium |
Publications (3)
Publication Number | Publication Date |
---|---|
EP3337135A1 true EP3337135A1 (en) | 2018-06-20 |
EP3337135A4 EP3337135A4 (en) | 2018-09-26 |
EP3337135B1 EP3337135B1 (en) | 2020-08-12 |
Family
ID=55333117
Family Applications (3)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
EP16850315.9A Active EP3337135B1 (en) | 2015-09-28 | 2016-09-26 | Acceleration management node, acceleration node, client, and method |
EP21217676.2A Withdrawn EP4040761A3 (en) | 2015-09-28 | 2016-09-26 | Acceleration management node, computer program product and non-transitory computer-readable medium |
EP20174899.3A Active EP3767912B1 (en) | 2015-09-28 | 2016-09-26 | Acceleration management methods |
Family Applications After (2)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
EP21217676.2A Withdrawn EP4040761A3 (en) | 2015-09-28 | 2016-09-26 | Acceleration management node, computer program product and non-transitory computer-readable medium |
EP20174899.3A Active EP3767912B1 (en) | 2015-09-28 | 2016-09-26 | Acceleration management methods |
Country Status (4)
Country | Link |
---|---|
US (3) | US10628190B2 (en) |
EP (3) | EP3337135B1 (en) |
CN (2) | CN111865657B (en) |
WO (1) | WO2017054691A1 (en) |
Families Citing this family (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111865657B (en) | 2015-09-28 | 2022-01-11 | 华为技术有限公司 | Acceleration management node, acceleration node, client and method |
CN105893036B (en) * | 2016-03-30 | 2019-01-29 | 清华大学 | A kind of Campatible accelerator extended method of embedded system |
CN107493485B (en) * | 2016-06-13 | 2021-11-05 | 中兴通讯股份有限公司 | Resource control method and device and IPTV server |
CN105979007B (en) * | 2016-07-04 | 2020-06-02 | 华为技术有限公司 | Method and device for accelerating resource processing and network function virtualization system |
CN106412063B (en) * | 2016-09-29 | 2019-08-13 | 赛尔网络有限公司 | CDN node detection and resource scheduling system and method in education network |
CN113328874B (en) * | 2016-11-30 | 2022-11-08 | 华为技术有限公司 | Data acceleration method, device and system applied to NFV system |
CN107436798A (en) * | 2017-08-15 | 2017-12-05 | 深信服科技股份有限公司 | A kind of process access method and device based on NUMA node |
US20190044809A1 (en) * | 2017-08-30 | 2019-02-07 | Intel Corporation | Technologies for managing a flexible host interface of a network interface controller |
CN108449273A (en) * | 2018-01-25 | 2018-08-24 | 上海连尚网络科技有限公司 | A kind of network accelerating method and system |
CN112041817A (en) * | 2018-05-08 | 2020-12-04 | 瑞典爱立信有限公司 | Method and node for managing requests for hardware acceleration by means of an accelerator device |
CN111625585B (en) * | 2020-05-22 | 2021-08-31 | 中科驭数(北京)科技有限公司 | Access method, device, host and storage medium of hardware acceleration database |
CN113014509B (en) * | 2021-05-26 | 2021-09-17 | 腾讯科技(深圳)有限公司 | Application program acceleration method and device |
CN117632457A (en) * | 2022-08-15 | 2024-03-01 | 华为技术有限公司 | Method and related device for scheduling accelerator |
Family Cites Families (21)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20050276413A1 (en) * | 2004-06-14 | 2005-12-15 | Raja Neogi | Method and apparatus to manage heterogeneous cryptographic operations |
US8434087B2 (en) * | 2008-08-29 | 2013-04-30 | International Business Machines Corporation | Distributed acceleration devices management for streams processing |
EP2442228A1 (en) * | 2010-10-13 | 2012-04-18 | Thomas Lippert | A computer cluster arrangement for processing a computaton task and method for operation thereof |
CN102685162B (en) | 2011-03-11 | 2014-12-03 | 中国电信股份有限公司 | Cloud computing acceleration method and system |
CN102148759A (en) * | 2011-04-01 | 2011-08-10 | 许旭 | Method for saving export bandwidth of backbone network by cache acceleration system |
US9167501B2 (en) * | 2011-08-29 | 2015-10-20 | Telefonaktiebolaget L M Ericsson (Publ) | Implementing a 3G packet core in a cloud computer with openflow data and control planes |
KR101861742B1 (en) * | 2011-08-30 | 2018-05-30 | 삼성전자주식회사 | Data processing system and method for switching between heterogeneous accelerators |
CN102650950B (en) * | 2012-04-10 | 2015-04-15 | 南京航空航天大学 | Platform architecture supporting multi-GPU (Graphics Processing Unit) virtualization and work method of platform architecture |
CN104348677A (en) * | 2013-08-05 | 2015-02-11 | 华为技术有限公司 | Deep packet inspection method and equipment and coprocessor |
CN103458029A (en) * | 2013-09-02 | 2013-12-18 | 百度在线网络技术(北京)有限公司 | Method, system and device for accelerating downloading through browser |
CN103596066B (en) * | 2013-11-28 | 2017-02-15 | 中国联合网络通信集团有限公司 | Method and device for data processing |
CN104951353B (en) | 2014-03-28 | 2018-09-21 | 华为技术有限公司 | It is a kind of to realize the method and device for accelerating processing to VNF |
CN104050045B (en) * | 2014-06-27 | 2017-06-27 | 华为技术有限公司 | Virtual resource allocation method and device based on disk I/O |
CN104102546B (en) * | 2014-07-23 | 2018-02-02 | 浪潮(北京)电子信息产业有限公司 | A kind of method and system for realizing CPU and GPU load balancing |
CN104503728B (en) | 2015-01-04 | 2017-11-24 | 华为技术有限公司 | A kind of hardware accelerator and chip |
CN104657308A (en) * | 2015-03-04 | 2015-05-27 | 浪潮电子信息产业股份有限公司 | Method for realizing server hardware acceleration by using FPGA (field programmable gate array) |
CN104765613B (en) * | 2015-04-21 | 2017-09-12 | 华中科技大学 | Towards the optimization method of tasks in parallel programming model under a kind of virtualized environment |
CN104899085B (en) | 2015-05-29 | 2018-06-26 | 华为技术有限公司 | A kind of data processing method and device |
US20160379109A1 (en) * | 2015-06-29 | 2016-12-29 | Microsoft Technology Licensing, Llc | Convolutional neural networks on hardware accelerators |
US10540588B2 (en) * | 2015-06-29 | 2020-01-21 | Microsoft Technology Licensing, Llc | Deep neural network processing on hardware accelerators with stacked memory |
CN111865657B (en) | 2015-09-28 | 2022-01-11 | 华为技术有限公司 | Acceleration management node, acceleration node, client and method |
-
2015
- 2015-09-28 CN CN202010506699.4A patent/CN111865657B/en active Active
- 2015-09-28 CN CN201510628762.0A patent/CN105357258B/en active Active
-
2016
- 2016-09-26 EP EP16850315.9A patent/EP3337135B1/en active Active
- 2016-09-26 EP EP21217676.2A patent/EP4040761A3/en not_active Withdrawn
- 2016-09-26 EP EP20174899.3A patent/EP3767912B1/en active Active
- 2016-09-26 WO PCT/CN2016/100137 patent/WO2017054691A1/en active Application Filing
-
2018
- 2018-03-28 US US15/937,864 patent/US10628190B2/en active Active
-
2020
- 2020-04-18 US US16/852,408 patent/US11080076B2/en active Active
-
2021
- 2021-07-15 US US17/376,305 patent/US11579907B2/en active Active
Also Published As
Publication number | Publication date |
---|---|
EP3337135B1 (en) | 2020-08-12 |
WO2017054691A1 (en) | 2017-04-06 |
US20180217856A1 (en) | 2018-08-02 |
EP3767912A1 (en) | 2021-01-20 |
CN105357258B (en) | 2020-06-26 |
CN111865657B (en) | 2022-01-11 |
CN111865657A (en) | 2020-10-30 |
EP3767912B1 (en) | 2022-06-22 |
EP4040761A2 (en) | 2022-08-10 |
US11579907B2 (en) | 2023-02-14 |
US20210342170A1 (en) | 2021-11-04 |
US10628190B2 (en) | 2020-04-21 |
EP3337135A4 (en) | 2018-09-26 |
US20200293345A1 (en) | 2020-09-17 |
US11080076B2 (en) | 2021-08-03 |
EP4040761A3 (en) | 2022-08-24 |
CN105357258A (en) | 2016-02-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US11579907B2 (en) | Acceleration management node, acceleration node, client, and method | |
US10698717B2 (en) | Accelerator virtualization method and apparatus, and centralized resource manager | |
US11411974B2 (en) | Applying policies to APIs for service graph | |
CN107534570B (en) | Computer system, method and medium for virtualized network function monitoring | |
US9639402B2 (en) | Systems and methods for automatic hardware provisioning based on application characteristics | |
US11025732B2 (en) | Method and apparatus to perform user authentication during cloud provider sessions | |
US10693860B2 (en) | RDP proxy support in presence of RDP server farm with session directory or broker | |
US9104501B2 (en) | Preparing parallel tasks to use a synchronization register | |
US20200319983A1 (en) | Redundancy Method, Device, and System | |
EP3584998A1 (en) | Method for virtual machine capacity expansion and reduction and virtual management device | |
US10686884B2 (en) | Method for managing sessions using web sockets | |
US8458702B1 (en) | Method for implementing user space up-calls on java virtual machine before/after garbage collection | |
CN114237937A (en) | Multithreading data transmission method and device | |
US10397071B2 (en) | Automated deployment of cloud-hosted, distributed network monitoring agents | |
WO2018137363A1 (en) | Method and apparatus for adjusting acceleration capability of virtual machine | |
EP3754499A1 (en) | Generating configuration templates for application delivery control | |
US10313429B2 (en) | Distributed resource management method and system | |
US9626444B2 (en) | Continuously blocking query result data for a remote query | |
US11614957B1 (en) | Native-hypervisor based on-demand code execution system | |
US10374893B1 (en) | Reactive non-blocking input and output for target device communication | |
CN113328874B (en) | Data acceleration method, device and system applied to NFV system | |
KR102146377B1 (en) | Method for monitoring virtual desktop and virtual host server in virtualization system and virtualization system thereof |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: THE INTERNATIONAL PUBLICATION HAS BEEN MADE |
|
PUAI | Public reference made under article 153(3) epc to a published international application that has entered the european phase |
Free format text: ORIGINAL CODE: 0009012 |
|
STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: REQUEST FOR EXAMINATION WAS MADE |
|
17P | Request for examination filed |
Effective date: 20180312 |
|
AK | Designated contracting states |
Kind code of ref document: A1 Designated state(s): AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR |
|
AX | Request for extension of the european patent |
Extension state: BA ME |
|
REG | Reference to a national code |
Ref country code: DE Ref legal event code: R079 Ref document number: 602016042020 Country of ref document: DE Free format text: PREVIOUS MAIN CLASS: H04L0029080000 Ipc: H04L0029060000 |
|
A4 | Supplementary search report drawn up and despatched |
Effective date: 20180827 |
|
RIC1 | Information provided on ipc code assigned before grant |
Ipc: H04L 29/06 20060101AFI20180821BHEP |
|
DAV | Request for validation of the european patent (deleted) | ||
DAX | Request for extension of the european patent (deleted) | ||
GRAP | Despatch of communication of intention to grant a patent |
Free format text: ORIGINAL CODE: EPIDOSNIGR1 |
|
STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: GRANT OF PATENT IS INTENDED |
|
INTG | Intention to grant announced |
Effective date: 20200305 |
|
GRAS | Grant fee paid |
Free format text: ORIGINAL CODE: EPIDOSNIGR3 |
|
GRAA | (expected) grant |
Free format text: ORIGINAL CODE: 0009210 |
|
STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: THE PATENT HAS BEEN GRANTED |
|
AK | Designated contracting states |
Kind code of ref document: B1 Designated state(s): AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR |
|
REG | Reference to a national code |
Ref country code: CH Ref legal event code: EP |
|
REG | Reference to a national code |
Ref country code: IE Ref legal event code: FG4D |
|
REG | Reference to a national code |
Ref country code: DE Ref legal event code: R096 Ref document number: 602016042020 Country of ref document: DE |
|
REG | Reference to a national code |
Ref country code: AT Ref legal event code: REF Ref document number: 1302681 Country of ref document: AT Kind code of ref document: T Effective date: 20200915 |
|
REG | Reference to a national code |
Ref country code: LT Ref legal event code: MG4D |
|
REG | Reference to a national code |
Ref country code: NL Ref legal event code: MP Effective date: 20200812 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: LT Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20200812 Ref country code: HR Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20200812 Ref country code: GR Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20201113 Ref country code: FI Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20200812 Ref country code: SE Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20200812 Ref country code: NO Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20201112 Ref country code: BG Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20201112 |
|
REG | Reference to a national code |
Ref country code: AT Ref legal event code: MK05 Ref document number: 1302681 Country of ref document: AT Kind code of ref document: T Effective date: 20200812 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: NL Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20200812 Ref country code: RS Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20200812 Ref country code: LV Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20200812 Ref country code: PL Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20200812 Ref country code: IS Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20201212 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: RO Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20200812 Ref country code: DK Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20200812 Ref country code: CZ Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20200812 Ref country code: EE Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20200812 Ref country code: SM Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20200812 |
|
REG | Reference to a national code |
Ref country code: CH Ref legal event code: PL |
|
REG | Reference to a national code |
Ref country code: DE Ref legal event code: R097 Ref document number: 602016042020 Country of ref document: DE |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: MC Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20200812 Ref country code: AT Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20200812 Ref country code: AL Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20200812 Ref country code: ES Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20200812 |
|
PLBE | No opposition filed within time limit |
Free format text: ORIGINAL CODE: 0009261 |
|
REG | Reference to a national code |
Ref country code: BE Ref legal event code: MM Effective date: 20200930 |
|
STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: NO OPPOSITION FILED WITHIN TIME LIMIT |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: LU Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES Effective date: 20200926 Ref country code: SK Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20200812 |
|
26N | No opposition filed |
Effective date: 20210514 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: IT Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20200812 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: CH Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES Effective date: 20200930 Ref country code: BE Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES Effective date: 20200930 Ref country code: SI Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20200812 Ref country code: IE Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES Effective date: 20200926 Ref country code: LI Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES Effective date: 20200930 |
|
REG | Reference to a national code |
Ref country code: DE Ref legal event code: R079 Ref document number: 602016042020 Country of ref document: DE Free format text: PREVIOUS MAIN CLASS: H04L0029060000 Ipc: H04L0065000000 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: TR Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20200812 Ref country code: MT Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20200812 Ref country code: CY Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20200812 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: MK Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20200812 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: PT Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20200812 |
|
P01 | Opt-out of the competence of the unified patent court (upc) registered |
Effective date: 20230524 |
|
PGFP | Annual fee paid to national office [announced via postgrant information from national office to epo] |
Ref country code: GB Payment date: 20230803 Year of fee payment: 8 |
|
PGFP | Annual fee paid to national office [announced via postgrant information from national office to epo] |
Ref country code: FR Payment date: 20230808 Year of fee payment: 8 Ref country code: DE Payment date: 20230802 Year of fee payment: 8 |